NCA-GENM practice questions
NVIDIA · NCA-GENM · 300 questions
Validates foundational skills for designing, implementing, and managing AI systems that synthesize and interpret data across text, image, and audio modalities, covering multimodal experimentation, core ML knowledge, NVIDIA Riva speech pipelines, diffusion-based image generation, software development, performance optimization, and trustworthy AI.
This course contains the use of artificial intelligence.
About the NCA-GENM exam
- Exam fee
- $125 USD
- Time allowed
- 1 hour
- Questions
- 50-60
- Format
- Online, remotely proctored, multiple-choice
Exam details published by the vendor, checked 28 August 2026. Vendors change fees and formats without notice — confirm on the vendor's own page before you book.
Practice Quizzes
Test your knowledge with standard 20-question practice sets.
Quiz 1
Quiz 2
Quiz 3
Quiz 4
Quiz 5
Quiz 6
Quiz 7
Quiz 8
Quiz 9
Quiz 10
Quiz 11
Quiz 12
Quiz 13
Quiz 14
Quiz 15
Browse by Domain
Study specific topics at your own pace.
Experimentation · 75 questions
- A lace-making archive scans pattern cards whose optical readout is full of broken words. Before a still-image matcher runs, a volunteer wants a transformer to tidy that text. What is the soundest first experiment?
- A dye-garden pigment desk clips handwritten cards to swatch photos. One trial asks a transformer only to pull the botanical name; another asks it to invent a poetic label. Which trial is analysis rather than generation for cataloging?
- A granite-quarry visitor radio dumps raw speech to text full of slang and repeats. The team wants clean place names before a photo-of-the-face lookup. How should they frame the transformer step?
- A paper-mill watermark lab already has a working still-image encoder. Staff want to try three transformer wordings for the accompanying note and leave the encoder untouched. Why is that a controlled multimodal experiment?
- A marionette-theater props closet pairs each puppet photo with a card. After a trial, every answer quotes the card and never mentions a missing string that is obvious in the photo. What should the next experiment test?
- A salt-pan field station wants to predict overnight brine density from five numeric sensors beside a webcam used only for later inspection. A contractor says to run a transformer on the log sentences. What is the best response?
- A cider-press tasting room dislikes a generated pour shot. One intern rewrites the accompanying sentence; another reruns the image generator from new noise. Which change is the text-arm experiment?
- A smokehouse batch desk records a spoken walk-through and snaps three rack photos. Staff ask a transformer to turn the transcript into a one-line batch ticket a later image checker can read. What kind of experiment is that?
- A clocktower carillon office dumps a weekend of visitor comments next to belfry photos. Staff want a transformer only to flag comments that mention a cracked louver so those frames stay in the keep pile. How should the experiment be framed?
- A ropewalk sail-loft captioner has a frozen image encoder. Staff can rewrite instructions, attach a small extra module on the language side, or retrain the whole text stack on a few dozen loft notes. Which framing best describes these choices?
- A funicular ticket window’s speech recognizer dumps a raw line. Staff then ask a pretrained token classifier to mark station names and car numbers so a map still can highlight the right platform photo. What is the text-arm experiment?
- An orchard packing-shed pins a one-line note to each crate photo. A pretrained text classifier should send “bruise” notes to the cull album and “blush” notes to the gift album. What is the right first experiment when notes are present?
- A library oral-history booth has hour-long transcripts next to a tray of portraits. Staff want a pretrained summarizer so a curator reads a short digest, then opens only the matching stills. What should the experiment evaluate?
- A pottery kiln glaze trial stores each firing as a card plus a tile photo. A volunteer types “which firing crawled?” and a pretrained question-answerer must point at the right card so the matching photo can be pulled. How should the experiment be scored?
- A cheese-aging cave has only forty labeled rind notes next to cave photos held fixed. One trial uses a pretrained token classifier as-is; another trains a small tagger from random weights on those forty notes. Which experiment is more sensible?
- A tide-mill visitor kiosk can speak a two-sentence summary or display the entire millwright transcript next to a gear photo. The team runs both and scores whether visitors find the right gear. What is the right success measure?
- A night-market voice stall files complaints that mix speech-to-text with a stall photo. Staff can try a pretrained classifier with no local labels, paste six examples into the instruction, or attach a small adapter on the text side. How should those options be treated?
- A boardwalk fortune booth records a short crowd clip and sends it through an NVIDIA AI Blueprint that uses VIA. Staff then type questions about who approached the booth. How should those typed questions be treated?
- An aquarium dive-bell office tags species names in a transcript, then a speech synthesizer reads the tags aloud to a gallery. Visitors hear the wrong Latin name even though the photo wall is correct. Where should the next single-stage experiment look first?
- A wind-farm blade-photo desk has inspector notes and stills. Leadership wants “something with a pretrained language model.” If the real need is to pull the matching still for a typed fault question, which text-arm trial should they run?
- A tram-depot wayfinding booth keeps hearing “car barn” as “carbon.” Staff can add a small site vocabulary on Riva ASR or leave the stock model. When is a customization experiment justified?
- After a tram booth adds “car barn” to Riva ASR, a supervisor wants a number before keeping the change. How should the team measure whether the customization helped?
- An ice-rink lost-child kiosk can speak slowly in a calm voice or quickly in a bright voice. The team keeps the recognizer fixed and only swaps Riva TTS settings, then asks parents which reply they understood. What is the experimental factor?
- A community-radio call-in booth can score Riva ASR by counting wrong words or by asking whether the producer still routed the caller to the right topic bin. Which measure matches the booth’s real job?
- A harbor foghorn museum wants either live captions beside an exhibit or a nightly dump of the day’s docent talks. How should the team choose between streaming and batch Riva ASR trials?
- A ropewalk loft updates Riva ASR site vocabulary, microphone gain, and the language pack in one weekend, then reports that recognition “got better.” What is the soundest next experimental move?
- A glasshouse orchid desk renders the identical weekly care notice with two custom Riva TTS voices and plays both for interns. What success measure fits that TTS comparison?
- An observatory plate-archive intern wants to push a half-finished Riva ASR vocabulary to the public kiosk “to see what happens.” What should happen first?
- A quarry radio still mangles the phrase “lewis hole.” Which set of experimental levers is appropriate for an audio+text booth built on Riva?
- A clocktower office hears that “Riva is using the GPU,” and an intern offers to write a custom CUDA kernel “to make the experiment fair.” What is the correct associate response?
- A salt-pan hut’s outdoor mic clips every gust. After three Riva vocabulary trials, word error barely moves. What should the next experiment target?
- A marionette theater plans a later talking-face demo with ACE, Riva ASR/TTS, and Audio2Face. How should a successful Riva TTS customization be treated in today’s experiment plan?
- A night-market stall hears a shopper, decides what they asked, and speaks a reply. Which three experimental units map to that hear–understand–speak flow in a Riva conversational pipeline?
- An ice-rink kiosk answers “your skates are in locker twelve” when a parent asked about a missing hat, yet the spoken words are crisp and clear. Which stage should the next experiment inspect?
- A tide-mill kiosk keeps Riva ASR and TTS frozen and tries two different pretrained language heads on the same recognized lines. Why is that the right control?
- A funicular window can map recognized phrases to a tiny reply table or send each line through a pretrained language model, then speak. How should that comparison be designed?
- A tram-depot booth can wait for a slightly better transcript or speak a faster, rougher reply so riders catch the next car. What experimental choice does that pose?
- A glasshouse desk recognizes English questions correctly, but the spoken reply comes out in a language visitors do not speak. What kind of experiment does that call for?
- A harbor-museum team updates the recognizer vocabulary, the reply templates, and the TTS voice in one weekend and cannot tell what helped. How should the next sprint be redesigned?
- A marionette box office already speaks through a speaker. Leadership asks whether adding an ACE face driven by Audio2Face on the same Riva TTS would change visitor success. How should that request be framed?
- An observatory wants searchable spoken plate-talk and will never speak back. A vendor sells a full Riva ASR+NLP+TTS stack. What experiment scope fits the job?
- A cider-press room runs both a Riva booth and a pour-shot generator. An intern pastes the booth’s results into the same “quality” column as the pictures. What should happen?
- A boardwalk team can field either a spoken Q&A booth or a short-clip review that uses an NVIDIA AI Blueprint with VIA. How should they choose the experiment?
- A quarry radio’s Riva pipeline finally beats the word-error target on held-out calls. Ops asks how that same stack gets installed in a cluster the same way every time. What answers that ask?
- Festival week triples kiosk turns at the ice rink. The conversational model is unchanged; people are waiting. What is the right move?
- A tram depot’s Helm chart is slow to start, and an intern wants to redesign the cluster fabric and vGPU pools. What is the associate-level response?
- After a good lab run of a new Riva vocabulary, a foghorn museum wants 10% of kiosk pods on the new chart values and the rest on last week’s. What is that?
- A glasshouse kiosk chart is live. One dashboard shows pods ready; another still tracks transcription error on a weekly phrase list. How should those signals be treated?
- A night-market stall changes NLP replies every afternoon and also asks to “Helm it for scale.” What should happen with cluster packaging?
- An intern edits replica count and resource requests in the conversational Helm chart and claims they “invented a new Riva model.” How should that change be labeled?
- A town help-desk Helm chart lists a GPU for the NVIDIA Riva speech containers. A volunteer offers to write a custom CUDA kernel “so Kubernetes can see the device.” What should the associate keep in scope for this conversational experiment deploy?
- A paper-mill watermark intern is told the image generator starts from a random field and removes noise in steps until a mark appears. How should the candidate restate that process?
- A dye-garden desk renders the same pigment card at several denoising-step counts. Smear fades, then further steps barely change the card. What should the quality experiment conclude about step count?
- A quilt-museum scanning room asks whether a diffusion quality trial may begin from a blank noisy field or must seed the generator with a real quilt photograph. What is the official starting point for such a generation experiment?
- A high-resolution lace-scan project hears that some generators work in a compressed stage and then decode to pixels. How should the associate use that two-stage idea in a quality experiment?
- A cider-press marketing pair wants a fair test of whether a longer denoise looks better. How should they control the experiment?
- A wind-farm still keeps showing a whole turbine when the note asked for a close trailing-edge chip. Doubling denoising steps only makes the tower sharper. What should the associate do next?
- A pottery glaze desk has two diffusion recipes and wants to compare them fairly. What must stay shared so the comparison is about the denoiser?
- A generated orchid card looks crisp but shows the wrong species color next to the Latin line. How should the associate choose a quality measure for the claim “matches the sentence”?
- A smokehouse wants tomorrow’s product cards to look like this smokehouse, not a generic shed. Which three image+text experiment levers should the associate compare?
- A clocktower postcard trial cannot be repeated: each run looks different, so staff cannot tell whether a sharper bell came from more steps or from luck. What control should the associate add?
- One observatory-plate render still looks speckled; another is waxy and loses a hairline crack that mattered. How should the associate read those diffusion-quality outcomes?
- A lace archive’s new denoiser wins a “looks like a real card” check, but a later text matcher still retrieves the wrong pattern name. What should the associate design next?
- A quilt-museum intern asks why changing the sentence changes the stitches in the render. What is the context embedding doing?
- A dye-garden wants “weld yellow on linen, shade, no flower,” but the first render shows a bouquet. What is the official test-and-refine experiment?
- A cider-press card must be generated from an English line. Which official path turns that line into a visual condition for the image experiment?
- A wind-farm render keeps adding a whole tower behind a close-up chip. Staff compare adding a second context that rules the tower out versus only rewriting the positive line. What is being tested?
- A pottery desk embeds “ash glaze, iron speck” and “celadon, quiet rim,” then runs both through the same diffusion recipe. What must stay shared so the factor is the context embedding?
- A paper-mill mark follows the sentence loosely at a weak text condition and looks harsh and over-literal at a very strong one. What associate experiment fits?
- A smokehouse wants house style on future cards and can keep iterating sentences, attach a small adapter, or fine-tune the whole generator. How should the associate choose?
- A boardwalk clip runs through an NVIDIA AI Blueprint with VIA; the first text query returns a useless recap. Staff refine the query and compare the new recap to a human note. What kind of experiment is this?
- A harbor museum claims “clearer context helps everything” and runs a tighter sentence on the postcard generator and a tighter reply template on the Riva booth. How should metrics be logged?
- After a dozen wording trials, a lace card still cannot show a rare stitch the generator never saw. What should the associate do?
- An automatic check says a kiln-tile render matches the card, but three potters say the pool is the wrong green. How should the associate treat those readouts?
- A quilt room freezes the generator weights and spends a week only on embeddings and seeds. Why defend that freeze?
Core Machine Learning and AI Knowledge · 60 questions
- A cooperage trains a network on stave photos paired with stave notes. An apprentice says the run succeeded because the preview “looks nice.” What quantity actually drives the weight updates during training?
- A lighthouse lens room walks through one training step on a cracked-prism still plus a short keeper note. Which statement correctly separates the two halves of that step?
- A honey-house desk trains on comb photos with harvest notes. One volunteer calls every single comb “an epoch,” and another calls a shuffled tray of sixteen pairs “the whole dataset.” Which assignment is correct?
- A milliner’s loft fits every training still and label perfectly, then misses next season’s new hat blocks. What pattern does that describe?
- A blacksmith shop trains a tiny network on quench stills plus temperature notes, and both the training loss and the held-out loss stay high. What does that pattern indicate?
- A rum rickhouse intern wants to tune every knob on the same barrel-end photos that will later be called “the test.” Why keep training, validation, and a final untouched split separate?
- A canal-lock keeper’s training curve either jumps wildly after each update or barely descends over many passes. What learning-rate idea best explains those two behaviors?
- A stained-glass studio watches training loss keep falling while the held-out loss turns upward. What training-loop idea fits ending the run when the unused split no longer improves?
- A charcoal-kiln camp stacks a very deep stills-and-notes network. In one recipe the error signal becomes tiny after a few layers; in another the numbers explode. What failure modes are those?
- A letterpress shop is ready to train on plate photos plus ink notes. A volunteer offers a hand-written update loop on a spreadsheet. What is the appropriate associate-level training tool?
- A sled-dog kennel hears that training the photo-plus-log model “uses the GPU.” An intern offers to write a custom CUDA kernel “so training counts.” What is the correct associate stance?
- A seed-vault has a few hundred packet photos with species tags and many more untagged packets. How do supervised labeled-pair training and a self-supervised pretext differ?
- A tannery hide loft must choose an architecture that reads grain photos and later a short grade note. Which pairing best matches common architecture strengths?
- A violin shop wants one network that compresses a belly photo into a code, one that writes a condition note from that code, and one that does both. Which role names match those jobs?
- A bookbindery stacks many layers on cover stills plus binder notes and early layers stop receiving a useful error signal. Which architectural idea adds the block’s input back to its output?
- An ice-cutter pond house needs English lines and pond stills to live in one comparable space. How does CLIP’s dual-encoder contrastive design work?
- A hop-kiln office hears that pictures are grown by starting from noise and removing it in stages. Which generative architecture is that?
- A hop-kiln diagram of a diffusion model shows a shrinking path, a growing path, and bridges between matching sizes. What common denoiser backbone is that?
- A map-archive wants the same transformer family that reads catalog sentences to also read a sheet by cutting it into patches. What architecture idea is that?
- An organ-pipe loft can wire every layer strictly after the last, or allow a parallel / skip graph so air and scale notes meet later. What contrast is that?
- A telescope-grinding shop hears that high-resolution synthesis often works in a compressed stage and then decodes to pixels. What architecture awareness does that describe?
- A carousel restoration barn wants a talking figure. Official additional material points to ACE with Riva speech, Audio2Face, and a NeMo LLM. What kind of architecture is that stack?
- A spice-mill captioner asks what is actually stacked inside the language model that will later sit beside stills. What is the reusable transformer unit?
- A cobbler bench places a care sentence next to a sole photo. Before the language stack can run on that sentence, what must happen to the text?
- A mustard-mill note says “add seed then grind,” which means the opposite if the words swap. Why does a transformer need positional information?
- A city flax-retting pond office needs three text jobs: a stack that only reads a note, a stack that only writes a new note, and a stack that reads one note then writes another. Which transformer layouts match those three roles?
- A municipal pigeon-loft log runs for many pages. Staff ask why a modern language model uses a transformer instead of stepping through the log with a long recurrent chain. What is the architectural reason?
- A wool-carding mill is offered “a modern pretrained language model” to sit beside fleece stills in a multimodal pilot. How should the team describe what that model is?
- A municipal kite-maker loft wants the language stack to look at sail-cloth stills while it writes repair notes. Which transformer mechanism hooks another modality into the text generation path?
- A mushroom-cave contractor starts listing prompt catalogs, retrieval chains, and text-alignment recipes from a text-only associate exam. For this multimodal exam, how should transformers be treated?
- A net-mending shed’s care manual is longer than the language stack can hold at once. What does the context window represent for that transformer LLM?
- A brick-kiln yard hears, “We bought a transformer, so it already sees firings and hears the roar.” What is the correct response?
- A copper-roofing shop has JPEG stills of seams ready for a vision model. What readiness step makes those stills neural-network ready?
- A linotype shop’s repair sentence must be multiplied by weight matrices inside a language model. What conversion makes that text neural-network ready?
- A soap-boiling shed records the kettle and a volunteer passes only the recording’s filename string into a speech model. What must happen before the model can run?
- A dune-grass nursery records a short planting clip and will later ask a typed question about it. Official materials allow video through an NVIDIA AI Blueprint with VIA. At data-readiness depth, how should that video be treated?
- A wax-seal archive has a seal photo in pixels and a motto in tokens. What readiness goal must both sides meet before a joint model can score a match?
- A piano-action shop’s parts catalog is a scanned PDF page. Which readiness step turns that page into text a network can use?
- A pearl-boat locker mixes 8 kHz radio clips with 48 kHz deck recordings and stills of three different sizes. Why is this collection not yet network-ready?
- An ice-fishing shack tags each hole photo “safe,” “thin,” or “open water.” How should those categorical tags be made neural-network ready for a training loss?
- A thatcher’s yard has notes of three words and notes of three pages next to roof stills. Why must variable-length text or audio be padded or packed before a batched forward pass?
- A silversmith dumps whole file blobs—PDF bytes, WAV bytes, camera bytes—into one array and calls it multimodal input. What is wrong with that approach?
- A bone-china studio concatenates raw pixel values next to raw token ids and feeds that concat into one stack. Which fusion type is that?
- A basket-weaving shed runs a stills-only grader and a notes-only grader, then averages or votes at the end. Which fusion type is that?
- A felt-hat factory runs a vision encoder and a text encoder to mid-level codes, then lets those codes interact—often by cross-attention—before a joint head. Which fusion type is that?
- A windmill flour loft concatenates millions of pixel numbers with a twelve-token note, and the stills drown the sentence. What known early-fusion weakness does that illustrate?
- An ink-making shop already trusts a stills grader and a smell-note grader and only needs a final mix. Updating one specialist must not retrain the other. Which fusion choice best fits?
- A coral-nursery log only makes sense if a bleaching patch in the photo is tied to a phrase in the diver note before any decision. Which fusion type should the team pick?
- A tea-tasting room has a spoken comment, a liquor photo, and a tasting slip. As a design concept, how do the three fusion types join those modalities?
- A pulley-block shop intern calls “late fusion” the name of a sharper picture. What correction is needed?
- A typewriter-repair bench claims “late fusion of audio and stills,” but the audio was never turned into features. What is wrong?
- An ice-house must choose a fusion type. Leadership either wants to “swap the stills model next month without touching text” or needs “the caption to point at a region.” How should those constraints map?
- A ferry-ticket office has a still of a pass, a spoken request, and a typed note. Something must decide which encoder each stream enters and when those streams may meet. What is that routing of data types called?
- A sled-maker loft wants the system to greet, look up a runner part, speak a reply, then optionally move a face. What kind of orchestration is that sequencing of jobs?
- An ice-climbing hut sometimes has only a spoken condition report and sometimes only a slope photo. A router sends the available type to the matching encoder. What is that pattern?
- A calliope workshop script says: acknowledge the visitor, fetch the tune card, then speak the title. Even if every step uses the same text encoder, what is this script?
- A lantern-maker intern points at an intermediate concat layer and says, “That is the agent.” What is the correct distinction?
- A canal-barge office has a planner that picks “ask a question” versus “show a still.” Someone files the planner under “late fusion.” What should happen?
- A tidal-pool lab uses an NVIDIA AI Blueprint with VIA on a short clip plus a typed question. At Domain 2 depth, how should that situation be read?
- A bell-foundry visitor desk can stream speech and drive a face (ACE / Riva / Audio2Face / NeMo LLM from official additional material) and can also follow a “listen, answer, smile” script. How should those concerns be split?
Multimodal Data · 45 questions
- A maple-sugar shack wants NVIDIA Riva ASR to recognize local terms such as “spile,” “niter,” and “hydrometer.” Staff prepare a word list plus example utterances. How should that material be treated for Riva ASR customization?
- A locksmith bench records spoken key-cut requests and types what was said for each clip. What is the correct unit of ASR customization data?
- An accordion-repair loft dumps a week of bench recordings with no typed lines and labels the dump “the Riva set.” What should happen before treating it as ASR customization data?
- A weathervane loft wants NVIDIA Riva TTS to read care notices aloud. A volunteer points at a pile of vane photographs as “the TTS data.” What should the team request instead?
- A ship-caulker records a long take that includes the phrases “oakum, then pitch.” One annotator marks when each phrase starts; another writes a single mood note on the whole file. Which annotation helps the speech recognizer learn from the clip?
- A parchment studio must choose whether to record its Riva ASR set over the scraping bench or in a quiet side room. The finished kiosk will listen on the busy floor. How should the recording-condition choice be framed?
- An indigo-vat floor has one talkative dyer and six quieter helpers, but the first ASR draft corpus contains only the dyer. What should the team flag before customizing Riva?
- A clay-pipe kiln writes how specialty terms such as “sagger” and “waster” should be pronounced for NVIDIA Riva ASR. How should that pronunciation lexicon be classified?
- A paper-marbling shop may later add a talking face using ACE with Riva and Audio2Face from official additional materials. What is the smart Domain 3 move for today’s audio–transcript pairs?
- A sign-painter kiosk must store what was heard, what was meant, and what should be spoken back for each customer turn. How should that record be treated as conversational pipeline data?
- A dry-stone yard has ASR transcripts and intent tags but no spoken-back lines for the kiosk. What does that mean for the conversational pipeline corpus?
- A millstone loft wants NLP to route “dress the stone” versus “buy flour,” but someone starts tagging hopper photographs instead. Where should the intent annotation go?
- A farrier stall’s ASR set is English, the intent table mixes languages, and the TTS scripts are written in yet another language. What should the candidate conclude about the three corpora?
- A gilder’s loft has leaf-and-frame photographs and no spoken turns. A contractor calls the album “the multimodal Riva set.” How should that claim be handled?
- A pewter bench records overlapping customers, and without turn boundaries the intent tagger cannot see where one request ends. What role do timestamps or turn segmentation play?
- A tinsmith row has a clean transcript, a blank intent field, and a spoken reply line left over from yesterday’s order. How should that example be diagnosed?
- A wainwright loft is only preserving axle talk for later reading and will not speak replies. How far should the speech data plan go?
- A malt-house floor snaps a floor-grain still and writes one English line that names what is in the frame. What is the CLIP data unit that describes?
- A sawmill office captions blade stills in a language the official CLIP skill path does not name for its English-prompt workflow. What should the team flag?
- A shingle-mill row shows a cedar shake, but the English line describes a slate roof from another bay. How should that CLIP row be treated?
- A lime-kiln camp has hundreds of pot stills and empty English caption fields. How should those rows be classified for CLIP?
- A coble yard has slogan lines such as “tarred bottom, oak strake” ready but no photographs. Can that slogan list be called a CLIP corpus?
- A last-maker can write what is actually on the bench or invent a poetic shop motto beside the still. Which English annotation belongs on a CLIP pair?
- A clog shop keeps stills in one drawer and English slogans in another with no shared identifier. What is required before those piles can become a CLIP paired corpus?
- A reed-organ loft can caption a still “walnut case, two manuals” or paste a three-page parts inventory. How should caption length and focus be treated?
- A rush-light shop intern scores caption–still rows with a spoken-word error count borrowed from the Riva booth. How should pairing quality for those rows be judged?
- A horn-button works has a still, a care sentence, and a tap-test clip that might belong to three different buttons. What must be true before anyone discusses early, late, or intermediate fusion on those files?
- A pin-mill row shows a brass pin in the still, but the audio clip is someone describing a steel coil from the next machine. What data problem should be marked?
- A needle-works row is planned as still + note + tap sound, but the microphone failed and no audio was captured. How should that row be treated in the multimodal corpus?
- A thimble shop loses the photo but still has the care sentence and a tiny click. Late fusion could run the remaining specialists; early concat cannot invent pixels. What should guide the next step?
- A fulling-mill intern concatenates a 4K still next to a three-word note and a 30-second clip with no shared clock or size layout. What should be flagged before calling the row early-fusion ready?
- A print-works desk wants a stills grader and a notes grader to vote at decision time, but the stills grader’s rows have no pictures. Can this yet be called a late-fusion dataset?
- A litho-stone loft plans to meet vision codes and text codes halfway for intermediate fusion. What data requirement comes first?
- A copperplate shop needs a label that is true only of the pair—this motto belongs on this plate—not of either file alone. Where does that annotation belong?
- A woodcut shop claims “late fusion of scent and stills” but never recorded any scent. What is the sound data response?
- An etching-bath intern wants to add a fourth unnamed product stream to every paired row “because multimodal means everything.” What scope should the corpus keep?
- An enamel kiln records a short firing clip and a typed question for an NVIDIA AI Blueprint with VIA. How should that material be treated as data?
- A wrought-iron shop sometimes has a spoken request and sometimes only a scroll still. What is the data side of modality orchestration for today’s example?
- A saddle-tree loft scripts greet → look up a tree → speak a reply, but the lookup table is empty. What should be named as the problem?
- A bridle loft planned still+speech rows, but the camera failed. How should today’s example be handled on the data side?
- A kelp-drying rack wants the system to “fetch the grade, then speak it,” but no grade table was built. Where should work stop?
- An eel-weir watch uses an NVIDIA AI Blueprint with VIA on a short clip plus a typed question. What are the data arms being orchestrated?
- An oyster-rake shed may later drive a face with ACE / Riva / Audio2Face. What data ask is appropriate at awareness depth?
- An ice-cutter pond office writes a polished greet / ask / show script but has no stills or captions. How should that script be filed?
- A packet-boat office argues about the next scripted agent action before anyone records the passenger still or the spoken request. What should happen first?
Software Development · 45 questions
- A mask-carver’s first app will either load stills of unfinished faces or stream a spoken note. What should decide which encoder module to import?
- An umbrella loft needs English lines and canopy stills to meet in one program. Which module should be plugged in as the official English-prompt image path?
- A harmonica works wants stills grown from noise inside the project. Which piece should be added as the callable image-generation module?
- A button mill already has a stills loader and needs a text arm for care sentences. How should that text module be added?
- A candle-dip shop must either classify a spoken note or write a spoken reply. What should guide which packaged architecture to import?
- A mosaic studio wants a talking face after Riva hears a visitor. From official additional material, which face module should be assembled on that path?
- A wheelwright intern offers to “make the vision stack faster by writing a kernel.” What is the associate implementation stance?
- A stone-carving yard’s first script feeds JPEG and WAV byte blobs straight into a model call. What implementation step is missing?
- An embroidery hoop loft is ready to write the first program that loads a stitch still and a care sentence together. How should that project start?
- An organ-pipe shop wants a single program that reads a pipe photo and a pitch note, then produces one output. How should that minimal multimodal path be sketched?
- A cork-cutting loft is past collecting turns and must stand up hear → understand → speak. Which services should be wired as the implemented conversational pipeline?
- A tile-glaze kiosk must accept a microphone stream and return spoken audio. Where do the official Riva stages sit in that served app?
- A harness shop wants a talking face. Official additional material names ACE with Riva ASR, Riva TTS, Audio2Face, and a NeMo LLM. What should the assemble choice be?
- A broom-maker’s first Python app works on one bench machine. An intern jumps straight to a Kubernetes chart. What should happen next?
- A lampwork bench asks what to do after the loaders exist. Which response fits associate implementation?
- A bead-furnace project “uses the GPU” through a modern deep learning framework. What does that imply for associate work?
- A looking-glass shop already has paired audio–transcript rows. Today they apply that set to a Riva ASR customization in the running app. How should that step be labeled?
- A spectacle-maker intern says “we picked the framework, so this is intermediate fusion.” What is the correct correction?
- A barometer shop’s generate button must produce a still from nothing on disk. How should that program path be implemented?
- A compass-maker exposes denoising steps as a call argument so a later experimenter can change it. What skill does that represent?
- A sundial yard hears an official two-stage high-resolution session: work in a compressed stage, then a second stage. At associate depth, how may that path be used?
- A lightning-rod loft loads yesterday’s still and calls it “generating from noise.” What must the noise-to-image skill require instead?
- A rain-gauge desk wants the same still tomorrow. How should reproducibility be implemented in the noise-to-image call?
- A bagpipe-reed shop wants the still faster. How should the generate path stay in associate scope?
- A hurdy-gurdy loft first implements “noise in, still out,” then will add an English prompt. How should the unconditioned path be assembled?
- A dulcimer loft’s first generate call returns a crooked instrument still. The encode-and-write path completed without error. How should that software milestone be treated?
- A lute-maker types “spruce top, ebony pegs” and needs a still from that English line. Which software sequence implements the official CLIP generate path?
- A harp loft is offered an unnamed third-party “prompt-to-still” product for English lines. For NCA-GENM associate implementation, what should the loft select?
- A cymbal-smith already paired English lines with shop stills for training. Today they call the generator from a brand-new sentence. How should that call be classified?
- A gong shop’s generate function must receive the text condition so the still matches the English line. At Domain 4 depth, what does the developer implement?
- A whistle-works demo keeps the same loaded CLIP and generator modules and only swaps the English string before rerunning. What kind of change is that?
- A flute-maker intern wires Riva ASR directly to “generate the picture” with no English text prompt. What is the correct Domain 4 correction?
- A fife shop has a short assembly clip plus a typed question and does not need a noise-to-image still. Which Domain 4 path should they implement?
- A clarinet-reed loft wants neighboring benches to request a still from an English line. At associate Domain 4 depth, what should they add?
- A trumpet shop has a short assembly clip and a typed question to answer about it. What does “customize a Blueprint with VIA” mean at Domain 4?
- A bugle desk must answer typed questions about a short drill clip. A contractor offers only a speech kiosk. Which Domain 4 choice fits the clip-plus-question job?
- A pretzel bakery sees an NVIDIA AI Blueprint with VIA that almost fits a proofing-room clip Q&A. Why customize that Blueprint instead of opening an empty modern-framework project?
- A coffee-roaster intern wants to “customize VIA” by bolting on unnamed extra sensors and a retrieval chain. What is the sound Domain 4 limit?
- A chocolate-temper bench asks what VIA stands for while customizing an NVIDIA AI Blueprint. The public cert materials print only the initials. What should the implementer do?
- A butter-churn loft already knows video and text are the two modalities. Today they must customize the Blueprint so the clip and question actually run. What domain step is that?
- A creamery already joined each clip to a typed question. Today they point that pair at a customized NVIDIA AI Blueprint with VIA. How should the pointing-and-customizing be labeled?
- A vinegar-cask room gets the VIA Blueprint answering on one machine. Someone opens a Kubernetes Helm chart next. What should Domain 4 finish first?
- A pickle-crock visitor desk wants a talking digital face, not clip search. Which Domain 4 assembly fits?
- A nailery intern opens a blank repository “to rebuild VIA from scratch.” What is the correct Domain 4 approach?
- A cabinetmaker’s customized Blueprint answers a typed question about a short joinery clip. How should that software outcome be counted?
Data Analysis and Visualization · 30 questions
- A municipal archive loft plots draft-note length against how many stills sit in each drawer and finds one drawer with long notes and almost no photos. What should that chart be read as?
- A civic magic-lantern loft bars slides with spoken lecture, slides with only a title card, and empty trays. Leadership asks for a fourth unnamed sensor so the bars look full. What should the steward do?
- A city hall photo tent pins a real captured plate beside a diffusion render of the same sitting. What is the best Domain 5 use of that side-by-side view?
- A visitor-center stereoscope parlor plots Riva ASR substitution errors against clock hour and sees a midday tour-group spike. How should that plot be interpreted?
- A zoetrope-style civic parlor stores tap-test audio clips beside drum stills. One intern pastes filenames on a wall; another opens a spectrogram next to the still. Which view better analyzes whether the audio arm is present and usable?
- A camera-obscura civic booth scans old handbills at mixed dpi and types what comes back. A scatter of resolution versus broken-word rate shows a clear knee. What is that plot for?
- A city kiosk log for a tintype-style photo booth is about 90% "how much" and only a handful of "retake." A pie of intents makes the skew obvious. How should that finding be treated?
- A cyanotype plant-book yard shades each catalog page by whether it has a plate, a caption, both, or neither. What does that page-wise view provide?
- An ambrotype-style civic studio flipped every sitting still "to grow the set." A wall of pairs now shows "scar on the left cheek" sitting on the right. What should that gallery analysis conclude?
- A praxinoscope booth ran an NVIDIA AI Blueprint with VIA on short clips plus typed questions. Stewards now table clip id, question, and answer to scan misses. How should that table be treated?
- A fletcher’s bench captions each still cock feather to the left of the nock. Someone mirrors every still to double the set. What should Domain 5 analysis conclude about that flip?
- A longbow tiller’s finish cards say linseed, not pitch-black. Heavy color jitter makes every tiller look tarred while the notes stay the same. What should the steward do?
- A Windsor-chair loft may later drive ACE / Audio2Face from official additional material. A helper speeds every spoken clip 1.4× for more audio. Why is that a poor Domain 5 augmentation choice?
- A bodger’s pole-lathe cards say knot on the grain. Random center-crops often drop the knot from the frame. What does Domain 5 analysis recommend?
- A trug-maker’s shed finds mild grain on harvest stills lifts holdout accuracy, while heavy noise erases the willow pattern though captions still say tight weave. Which trial matches the official augment-to-improve idea?
- A scythe-smith has twelve carefully paired quench stills and notes. An intern wants fifty random warps because augmentation always helps. What should Domain 5 analysis say?
- A brew-house mash floor flips stills of the tun and leaves tasting sentences that say left tap running unchanged. How should that one-arm augmentation be analyzed?
- A peat-cutter’s bog hut compares holdout scores before and after a mild stills-and-audio augmentation. Scores rose and pairs still match. How should that result be filed?
- A telegraph office has a week of typed fault notes clipped to insulator stills. A steward asks a transformer only to summarize themes so a chart of ice, wind, and vandalism can be drawn. What is that use?
- An hourglass shop’s care sentences hide shop words such as pontil and anneal. A transformer token-tags those terms, and a bar of tag counts shows anneal almost never sits next to a still. How should that chart be read?
- A sextant loft wants a one-page briefing of what the paired notes actually contain. One intern asks the transformer to invent lyrical captions; another asks it to report counts and contradictions. Which request matches generate-as-analysis?
- A coach-builder’s bench notes mix spellings of felloe. A transformer normalizes the word so a frequency chart stops splitting one concept. What is that step?
- A plough-wright contractor starts listing retrieval chains, prompt catalogs, and unnamed text-only products because a transformer is in use. What should the steward do for this multimodal corpus?
- A granary loft already filed every sack photo. Today they only want a transformer report on moisture comments so a steward can chart damp versus sound. What is the correct Domain 5 framing?
- A slate-pencil works has a bound PDF of old stock lists and a tray of pencil stills. What official Domain 5 path extracts the list text so it can be analyzed next to the stills?
- A chalk-pit office runs OCR across a hundred-page PDF and shades pages with high garbage rates. What is that page chart for?
- A marble-polish shed’s PDF names each slab on a line, and the camera tray uses the same slab numbers. After OCR, what Domain 5 step joins the document to the images for analysis?
- A flint-knapper’s PDF has a parts table on one page and a narrative on the next. Naive OCR dumps both as a word soup. What should analysis flag?
- A cameo-carver intern wants either to flatten PDF bytes into a training tensor or to start a Domain 1 wording trial on garbage text. What is the Domain 5 path first?
- A marquetry loft’s scanned PDF has clean type in the middle and pencil notes in the margins. OCR reads the type and drops the pencil. What analysis finding should the steward report?
Performance Optimization · 30 questions
- A city harpsichord workshop has only a few dozen labeled jack stills for a vision model. Training from random weights is too slow for the schedule. What should the candidate do?
- A municipal chronometer bench needs the visitor booth to recognize shop terms like “escapement” and “fusee.” Building a speech recognizer from silence would need a huge labeled set. What is the efficient path?
- A civic spyglass shop wants English lines to meet stills. One plan trains two encoders from nothing; another reuses contrastive language-image pretraining. Which start is the transfer-learning win?
- An amber-polish cooperative has forty labeled bead stills. A vendor insists thousands of new tags are required “or the net cannot start.” How should the candidate answer?
- A jet-carver’s single bench machine cannot finish a from-scratch vision run. Someone offers to write a CUDA kernel. What is the associate-level performance move?
- A mother-of-pearl inlay loft reuses a pretrained stills stack. How should they adapt it with less computation than updating every layer?
- A parquetry floor shop has a pretrained furniture-stills model. Where should transfer be expected to help, and where should it not?
- A treen-turner’s vision model is already adapted and accurate on one bench. An intern opens a Kubernetes values file “to do performance.” What should today’s job stay focused on?
- A chair-caning visitor desk may add a talking face later. Official extra materials name ACE with Riva ASR/TTS, Audio2Face, and a NeMo LLM. How should the speech path be handled for efficiency?
- A sieve-maker has a short assembly clip and a typed question and will customize an NVIDIA AI Blueprint with VIA. A volunteer starts a from-scratch video encoder. What should the candidate do?
- A bellows shop records a shop-only pressure trace that no pretrained stills or speech model has ever seen. What is the honest call?
- A coach-painter already has a pretrained stills model that nearly matches shop pigments. They can adapt a small last stage or retrain every weight. Which choice is the official less-computation path when the source is close?
- A still-room kiosk hears a visitor and should speak before they walk away. The steward times hear → understand → speak. What is that end-to-end wait called?
- A refectory visitor desk survives a feast-day line only if many short turns finish each minute. What should the candidate call that rate, and where should they look first?
- A chapter-house booth is fast after extra pods but still hears “misericord” as “mercy cord.” What should the candidate conclude?
- A lych-gate greeter is live on the conversational Helm chart. An intern wants to “drop in” an unnamed third-party server for speed. What should the candidate do?
- A porter’s lodge must answer live, while a night job captions a week of archive speech. How should targets be assigned?
- A toll-house intern wants a named quantization or precision product “because Performance Optimization.” That product is not on the official NCA-GENM skill list. What should the candidate keep?
- A coaching-inn yard’s Riva hear–understand–speak path works on one machine. Festival coaches arrive tomorrow. What is the official production-scale step?
- A posting-house desk keeps the same adapted models; only the line of riders grew. What should the candidate change?
- A gatehouse intern starts drawing spine fabrics and GPU virtualization “because Kubernetes.” Where should the candidate keep the work?
- A turnpike booth’s chart requests a GPU for Riva speech containers. A volunteer offers a custom kernel “to make the GPU count.” What is the correct stance?
- A falconry mews booth is slow in recognition, not in speech. Spoken replies are already instant. What should the candidate scale?
- A vestry desk intern edits replica count and resource requests and claims they “invented a new fusion type.” How should those edits be classified?
- A scriptorium kiosk still lacks paired transcripts (Domain 3 work). Ops wants the Helm chart “so it feels production.” What should happen first?
- A cloister booth already wired Riva ASR, NLP, and TTS in a framework project. Today pilgrims queue. What step is next?
- A buttery hatch already beat its word-error target in Domain 1. Meal hour now triples turns. What should the candidate do?
- A spring-house team adapted a new Riva vocabulary (transfer) and wants it on some pods first. How should they roll it out?
- A root-cellar tour chart is live. One board shows pod readiness and reply delay; another asks “is the voice a real steward?” Which board is Domain 6?
- A well-house customized an NVIDIA AI Blueprint with VIA and it answers on one bench. Someone opens a conversational Helm chart “because video should scale the same way.” What is the correct scope?
Trustworthy AI · 15 questions
- A portrait-miniature gallery hangs a diffusion render beside a real ivory miniature. Visitors treat the sharper frame as painted from life. What should the studio do for content authenticity?
- A wax-figure workshop plays Riva TTS that visitors hear as a living guide naming a real steward. What is the authenticity concern?
- A choir-loft kiosk drives an ACE talking face with Audio2Face while the voice was adapted from a chorister and the face is a stock figure. What should the candidate flag?
- A town-crier poster was grown through a two-stage latent high-resolution path and looks like a period broadside. What authenticity step still applies?
- A family-album scanner sends a short clip plus a question through an NVIDIA AI Blueprint with VIA. A relative asks whether the clip is the real picnic. How should the candidate treat that question?
- A silhouette parlor’s CLIP-steered generator passes a “looks crisp” check but invents a sitter who never visited. Can the studio file that under a trustworthy model?
- A cylinder-record parlor prints English-prompt stills from a CLIP path as if they were period sleeve photos. What provenance should the catalog card carry?
- A life-mask studio intern asks for an unnamed watermarking product “because Trustworthy AI.” That product is not on the official NCA-GENM materials. What should the candidate do instead?
- A wedding-portrait desk wants ACE / Audio2Face to replay a couple’s vows with a talking face. What must happen before faces and voices are reused that way?
- A civic-statue plaque generator, given “the foundry crew,” always renders the same body type and dress. How should a trustworthy-build review treat that pattern?
- A memorial-voice kiosk keeps raw visitor recordings so Riva can be adapted later, but families never agreed to a permanent voice file. What does a trustworthy build require?
- A carnival photobooth offers “hear the mayor say your name” by driving an ACE face from a public clip without the mayor’s consent. What is the trustworthy-AI judgment?
- A baptism-register scan project wants every face searchable from a typed name. What should the candidate ask before treating that as a default feature?
- A school-yearbook plate set pairs stills with sloppy captions that praise groups unevenly. A later CLIP-style generator repeats the slant. What should the candidate do?
- A mutoscope parlor wants a one-line practice for trustworthy multimodal models covering stills, speech, and a possible ACE face. A vendor pushes Guardrails, confidential GPUs, and servers from another exam. What summary fits this exam?
These questions are original practice material and are NOT actual exam questions or brain-dump content. All vendor marks are trademarks of their respective owners. This site is not affiliated with, endorsed by, or sponsored by the exam vendor.