--- title: "Industry AI Solutions | YPAI" url: https://ypai.ai/industry-solutions/ description: "AI data, evaluation and implementation shaped around the operating constraints of automotive, healthcare, finance, industrial, public-sector and education work." source: "src/copy/routes (route /industry-solutions/)" --- # Every industry breaks AI in a different place. Name the workflow. We set the route. > AI data, evaluation and implementation shaped around the operating constraints of automotive, healthcare, finance, industrial, public-sector and education work. We deliver AI Data & Evaluation and AI Implementation around the workflow you run and the constraint it runs under, whatever the industry. Below, what we produce, industry by industry. AI Data & Evaluation · AI Implementation · one engagement when the work needs both Two service lines, scoped around the same workflow. Data collection, sourcing, annotation, human review, quality assurance and model or workflow evaluation across speech, video, image, text and sensor data. Assistants, agents and automated workflows, designed, built and integrated into the systems and controls already in place. Either service line can stand alone. Where the work depends on both, YPAI scopes them as one engagement. - expert model evaluation · graders who know the domain - [AI companies and labs](https://ypai.ai/solutions/ai-companies/) - retrieval and document AI · your systems and controls - [Enterprise AI](https://ypai.ai/enterprise-services/) - learning and assessment · language work for providers - [Education](https://ypai.ai/solutions/education/) - EO/IR, geospatial and sensor data · specialist panels - [Defence and national security](https://ypai.ai/solutions/defence-national-security/) - egocentric video and demonstrations · task sequences and sensor sync - [Robotics and physical AI](https://ypai.ai/solutions/robotics/) - rights-cleared datasets · licensing that holds - [Datasets and licensing](https://ypai.ai/data-solutions/) - talking heads and lip sync · rights that hold - claims intake · a reviewable record - field imagery · crop-level ground truth ## We deliver into every industry. The industry is a label. The route is set by the workflow you run and the constraint it runs under. The industries below we show in depth. Every other sector runs on the same two service lines and the same delivery. Pick yours, and the page answers. ## Your model passed the benchmark. Production asks three questions the benchmark never did. The cabin, the ward and the counter are not in the training set. Every engagement here is specified against these three, and each is answered where the product runs. Answer the three for your workflow and the route sets itself. Where one cannot be answered in place, we say so at scoping. ### Was the data recorded where the product runs? - A corpus from a booth has never met the cabin. The model fails there first. - the course, in the learner's language ### Does the system respect who decides? - An answer from outside the approved sources is a control failure before it is a model failure. ### Was it proven in place before release? - Without an evaluation set from production, every change to the model ships on impression. ## Buy the data. Buy the system. Or both, from one team. - Collection, annotation, human review and evaluation sets recorded and graded where your product runs, in the languages it meets. - [AI Data & Evaluation](https://ypai.ai/data-solutions/) - Assistants, agents and document workflows built into the systems and controls already in place, with the record kept. - [AI Implementation](https://ypai.ai/enterprise-services/) ### Connected delivery - One team owns the system and the data it runs on. When the failure sits in the data, the people who built the system fix it. YPAI delivers for · automotive and mobility Below, the industries we show in depth. Every other sector runs on the same two lines and the same delivery. Name yours in the probe and the page answers. ## One specimen. Five modalities. One gate it has to pass. - [Speech data](https://ypai.ai/speech-data/) - [Video data](https://ypai.ai/video-data/) - [Physical AI data](https://ypai.ai/physical-ai-data/) - [Annotation](https://ypai.ai/data-solutions/annotation/) - [Evaluation](https://ypai.ai/data-solutions/evaluation/) - [Validation](https://ypai.ai/data-solutions/validation/) Each item is recorded where the product runs and leaves with a manifest you can verify against the batch. Review by people native to the language and qualified in the domain follows the agreed acceptance and sampling plan. - Reviewed by people who speak it - native and domain-qualified review on every batch - batch, rights and a hash for every item ## Cited, gated, chained. Edit one word in the source and the answer withdraws. - [Assistants and chatbots](https://ypai.ai/ai-assistants-chatbots/) - [Knowledge assistants](https://ypai.ai/enterprise-knowledge-assistants/) - [Document AI and workflow automation](https://ypai.ai/document-ai-workflow-automation/) - [Discovery and architecture](https://ypai.ai/ai-implementation/discovery-and-architecture/) - [Private enterprise deployment](https://ypai.ai/ai-implementation/private-enterprise-deployment/) Every answer is a passage from an approved source. Everything beyond that goes to a person. Each log line carries the hash of the one before it. - cited inline, refused when the evidence is missing - the same turns, with hard variants, form the release gate - every decision, model and policy version, on the log What we produce, industry by industry. ## The cabin · place the noise, play the utterance, watch the lattice run it on your own · moonshine tiny · in this tab load the model · about 28 MB model ready · runs on wasm in this tab cabin · the same recording under the noise runs in this tab · weights load from ypai.ai on demand the specimen is a synthesised voice run it on your own · electra small · in this tab load the model · about 14 MB Follow-up in six weeks for Kari Nordmann at Ullevål, repeat the bloods before the visit on 4 March. nothing to hold in this line names, places and organisations by the model · dates by rule ## The pixel mask · locked to the frame, measured in millimetres, snapped to the order inspection camera · line 4 · shift B · frame 0418 WO 88213 · snapped · coordinates inside station 4 no order · the mask is outside station 4 held · the mask does not lock run it on your own · minilm retrieval · electra reader · in this tab load the models · about 36 MB out of scope · no passage applies · routed to a named caseworker seal · sha-256(question ‖ span bytes ‖ version ‖ approved time) dense retrieval over the approved passages, then an extractive reader on the best one · the answer is a passage from the source - how long is a late application accepted - how do I appeal a decision - Applications for housing benefit are submitted through the municipal portal or on paper at the service centre. - The application deadline is the 25th of the month for benefit in the following month. - Applications are assessed against the regulation in force on the date of receipt. - Late applications are accepted up to 14 days after the deadline where the delay is documented. - Household income is assessed on the last three months of documented income - The benefit is paid on the 15th of each month to the account registered with the applicant. - A change of address must be reported within 14 days - A decision may be appealed within three weeks of the date the decision was received. run it on your own · electra reader · in this tab load the reader · about 14 MB reader ready · runs on wasm in this tab Application for credit facility, specimen. The applicant A. Lovelace requests a working-capital facility of NOK 240 000 and confirms the information given below. Registered address: 14 Verkstedgata, 0552 Oslo. Purpose, security and repayment terms as specified in annex A. Identity document issued 2021-05-02, validity confirmed against the register. The address on the identity document is 7 Løkkeveien, 4008 Stavanger. The applicant declares the information complete and correct. ## Who is the applicant? ## What is the registered address? ## What facility amount is requested? ## What is the address on the identity document? registered address matches the identity document · auto-approved for the next check registered address differs from the identity document · routed to review one question per field, answered as a span of the text · a rule compares the two addresses run it on your own · slimsam · in this tab load the model · about 12 MB embedding the frame · once, then every click is instant click a point in the frame · the model returns the mask a point prompt to a segmentation model · the mask it returns feeds the same millimetres, severity and work order ## When the failure sits in the data, the team that built the system fixes the data. A corpus that never met the cabin. A document type missing from the ground truth. An evaluation set that does not represent production. One team owns the system and the data it runs on, and either line can be bought alone. - Assistants, agents, document AI and workflow automation, delivered into your operating environment. - [The AI Implementation service line](https://ypai.ai/enterprise-services/) - Speech, video, image, text and sensor data recorded where your product runs, reviewed and evaluated through its lifecycle. - [The AI Data & Evaluation service line](https://ypai.ai/data-solutions/) ## Built and run from Norway, under the rules you already answer to. YPAI is a Norwegian company operating under GDPR. The system runs in your environment, inside the access boundaries you already enforce. Every engagement runs on Controlled Delivery, with acceptance criteria, records and gates you can inspect. - **Jurisdiction**: Norway, under GDPR - **Data residency**: EEA by default, defined per project - **Deployment**: Private and EEA, in your environment - **Processing terms**: Standard DPA, subprocessors and transfers defined per project - **Erasure**: 30-day end-of-contract erasure - **EU AI Act**: Article 10 documentation, produced as part of delivery - [Data residency in the EEA](https://ypai.ai/data-residency-eea/) - [AI security](https://ypai.ai/ai-security/) - [Provenance audit](https://ypai.ai/compliance/provenance-audit/) - [Ethical framework](https://ypai.ai/data-solutions/ethical-framework/) ## The rules come first. Three we bring to every first meeting. ### [GDPR and AI: Enterprise compliance requirements](https://ypai.ai/blog/compliance/gdpr-and-ai-articles-compliance/) GDPR applies directly to AI training data collection, model outputs, and automated decisions. What enterprise compliance officers must address in 2026. ### [EU AI Act Article 10: Data Governance Checklist](https://ypai.ai/blog/compliance/eu-ai-act-article-10-data-governance/) ### [CTO's Guide to Sovereign AI: Architecture & Costs](https://ypai.ai/blog/infrastructure/ctos-guide-sovereign-ai-architecture-costs/) ## Five questions we hear before scoping. Everything else is answered in scoping, with a named project lead. ### Is my industry not shown here? Every industry runs on the same two service lines and the same Controlled Delivery. The ones shown are where we can show the work in depth. Describe the workflow and the constraint, and a project lead sets the route. ### Can our data stay in the EEA, or inside our own environment? Yes. Processing, access, storage, review and delivery controls are designed around the data classification and your requirements, and defined before access begins. EEA-based, client-environment, private cloud, on-premises and hybrid configurations are assessed per use case. ### Do we have to buy both service lines? Either line stands alone. Where the work depends on both, YPAI scopes them as one engagement so the team that built the system also owns the data it runs on. - [The two service lines](https://ypai.ai/data-solutions/) ### What does a pilot demonstrate? Output quality, the proposed workflow, the review architecture, formats and the agreed acceptance criteria, against your requirements. Scope and commercial terms are agreed before the pilot starts, and production is a separate decision after review. - [How a pilot runs](https://ypai.ai/pilots/) ### Who reads what I write in the form? A named project lead. They read the workflow and the constraint, set the route to one line or both, and reply with feasibility and a first read on the constraint. ## What we deliver in your industry. The industries we show in depth. In each, the constraint that breaks generic AI, what we deliver against it, and the proof, settled. ### Your sector A workflow we have not drawn yet. That changes nothing about the offer. The same two service lines and the same Controlled Delivery apply. What changes is who reads your words first. the route, set by a person who has read your workflow - A named project lead reads the workflow and the constraint you describe. - They set the route to one service line, or one engagement across both. - You get a reply with feasibility and a first read on the constraint. named clients · automotive and mobility ## Cerence AI ## Hyundai ## BYD ## Honda ## Kia ## NIO ## Nexdata ## Is your workflow here? Every engagement starts as a pair: the workflow you run and the constraint it runs under. These are the pairs we have worked. Pick one and the page answers; if yours is missing, describe it and a person sets the route. - false accepts and rejects measured per market ## [Data recorded where the product runs, reviewed by people who speak it.](https://ypai.ai/data-solutions/) service line 01 · ai data & evaluation Collection, sourcing, annotation, human review, quality assurance and model or workflow evaluation across speech, video, image, text and sensor data. Consent and provenance documented on every item. in your industry it lands as - rights-clean clinical corpora with human review - specialist evaluation sets and document ground truth - annotated inspection and QC data from the running site - multilingual language data that serves every citizen - learning data reviewed by licensed educators and evaluated per subject - the corpus, evaluation set and annotation your constraint demands speech · video · image · text · 3D and sensor · annotation · evaluation · expert work · licensing · synthetic ### The corpus bench · one review gate, five modalities - follow-up in six weeks, repeat the bloods before the visit - transcribed · diarised · PII tokens held - inspection camera · line 4 · frame 0418 - tracked across 12 frames · class confirmed - cabin still · driver seat · night - mask locked to pixels · seat position attached - The applicant requests a working-capital facility of NOK 240 000 secured on annex A. - entities extracted · reviewed against policy - lidar sweep · pallet · 2.4 m ## [Assistants and workflows built into the systems already running, with the record kept.](https://ypai.ai/ai-implementation/) service line 02 · ai implementation Assistants, agents, retrieval, document processing and workflow automation, designed, built and integrated into your operating environment and its controls. Evaluated, monitored and improved after go-live. - in-cabin voice workflows the data serves - workflow assistants that respect the rights question - reviewable document workflows and controlled internal assistants - operational workflows around inspection and QC - assessment support that leaves the decision with the educator - the assistant, agent or workflow your constraint allows assistants · agents · retrieval · document processing · automation · integration · governance · deployment ### The controlled workflow · approved sources, a person in the gate, a chained log #### What documents are needed for a working-capital facility? A completed application, a registered-address document and an identity document. Security as specified in annex A. #### Is the registered address verified? The registered address differs from the identity document. Held for review. #### Summarise the applicant's risk and recommend a decision. Out of scope. Routed to a named analyst. routed · analyst named on the engagement decision log · hash-chained · each line carries the previous hash genesis · model 2.4.1 · policy v4.2 #### Credit policy v4.2 · §3 · approved 2026-02-11 09:40 A completed application, a registered-address document and an identity document are required for a working-capital facility. Security as specified in annex A. source changed · answer withdrawn · re-approval required ## The same two service lines, in any industry. how we help · two service lines Industries differ in where AI breaks, not in what answers it. Every engagement draws on the same two lines, scoped around the workflow and its constraint, and runs on the same Controlled Delivery. - The corpus, evaluation set and annotation your constraint demands: recorded where the product runs, reviewed by people who speak the language, with consent and provenance documented. - The assistant, agent or document workflow built into the systems already running your industry, with the gates, records and evaluation that procurement and compliance ask for. Either line stands alone. Where the work depends on both, YPAI scopes them as one engagement. Both run on Controlled Delivery: scope and acceptance criteria before work begins, rights and provenance documented, data and systems isolated, human review and delivery records inspectable throughout. ## Where we have gone deepest. One showcase per industry. In each, the constraint that breaks generic AI in that industry, and the deliverable that answers it, working. ### Automotive & Mobility - booth accepted · cabin false reject - In-cabin voice workflows built on the corpus, wake word to intent, inside the vehicle's HMI. - Cabin-recorded corpora and evaluation sets, native speakers, seat and noise metadata on every turn. - in-cabin voice · real cabin noise - The cabin itself: noise, light, languages and real driver behaviour that rarely match a training set. - Voice that ships to a cabin is proven in one before release. We record, annotate and evaluate against your acceptance criteria, and build the in-cabin voice workflows the data serves. - cabin-proven corpora and evaluation sets · in-cabin voice workflows - [Automotive & Mobility at YPAI](https://ypai.ai/automotive/) - [In-cabin voice and speech workflows](https://ypai.ai/voice-recognition-for-automotive/) - [In-cabin speech and perception data](https://ypai.ai/solutions/automotive/) - booth corpus · accepted in evaluation - cabin at 110 km/h · false reject - recorded in the cabin · native speakers · accepted #### Cerence AI #### Hyundai #### BYD #### Honda #### Kia #### NIO #### Nexdata #### The same utterance, twice ### Healthcare - consent on · in corpus · rights attached - Documentation and workflow assistants that respect the rights object on every turn. - Rights-clean clinical speech and documentation, de-identified, reviewed by people qualified in the domain. - clinical documentation · lawful use only - The rights question: what documentation, speech and imagery a project may lawfully use. - Nothing enters a clinical corpus without the right to be there. We collect and review clinical speech and documentation under that question, and build the workflow assistants that respect it. - rights-clean clinical corpora with human review · workflow assistants - [Healthcare AI at YPAI](https://ypai.ai/solutions/healthcare/) - [Clinical AI systems](https://ypai.ai/solutions/healthcare/clinical-ai-systems/) - [Clinical data collection](https://ypai.ai/solutions/healthcare/custom-data-collection/) - [Clinical model evaluation](https://ypai.ai/solutions/healthcare/clinical-model-evaluation/) - consent recorded · capture path open - de-identified · review accepted · in corpus · rights attached - day 91 · expired under retention · tombstone retained #### Nothing exists before the switch mic off · capture path closed - , repeat the bloods before the visit on - **purpose**: model evaluation - **basis**: consent · art. 6(1)(a) - **retention**: 90 days from capture - **source**: this session · specimen de-identified · tokens held, not deleted into corpus · locked until review in corpus · batch 0412 · rights attached advance the clock to day 91 - expired under retention · content erased - tombstone retained · id 0412-0189 · content hash · erasure time ### Financial Services - address mismatch lit · routed to review - Document processing and controlled internal assistants that log input to output. - Extraction ground truth, exception taxonomies and evaluation sets built from your own cases. - document processing · a reviewable record - Risk and compliance cannot approve a workflow they cannot inspect. - A workflow risk can approve is one that keeps its record. We build document processing and controlled internal assistants that log input to output, evaluated against your policies and precision thresholds. - reviewable document workflows · evaluated internal assistants - [Financial services AI at YPAI](https://ypai.ai/financial-services-ai-solutions/) - [Controlled internal assistants](https://ypai.ai/enterprise-knowledge-assistants/) - input hashed · model and policy versions logged - registered address differs from the identity document - routed to human review · reviewer named on the engagement #### Input to output, on the record - application for credit facility · page 1 of 4 · specimen - The applicant requests a working-capital facility and confirms the information given below. - facility details · page 2 of 4 - Purpose, security and repayment terms as specified in annex A. - identity document, copy · page 3 of 4 - Issued 2021-05-02. Validity confirmed against the register. - declarations · page 4 of 4 - The applicant declares the information complete and correct. - **input**: applicant. sha-256 - **model**: applicant. extract 2.4.1 · specimen run - **check**: address. registered address differs from the identity document - **id**: idaddress. address on identity document held - **routed**: address. to human review · reviewer named on the engagement · 09:41:07 ### Industrial & Energy - severity 2 · MES matched · operator hold - Operational workflows around inspection and QC, matched to the MES work order. - Inspection video and QC imagery captured on the running line, annotated to the pixel. - site inspection video · systems already running - The site is already running. The data has to come from it, and the system has to fit what is there. - We capture and annotate inspection video and quality-control imagery on the working line, and build the operational workflows around the MES, the cameras and the shifts already in place. - annotated inspection and QC data · operational workflows - [Industrial & Energy at YPAI](https://ypai.ai/solutions/industrial-energy/) - [Robotics and industrial vision annotation](https://ypai.ai/annotation/robotics-industrial-vision-annotation-services/) - surface porosity · severity 2 · mask locked to the frame - matched to the MES work order #### One frame, classified against the line inspection camera · line 4 · shift B - **condition**: dust, vibration, sodium light - **system**: matched to the MES work order - **verdict**: held for operator confirmation ### Public Sector - sealed · verified · out of scope routed - Citizen and caseworker assistants that cite approved sources and route exceptions to a person. - Multilingual speech and text recorded by native speakers, evaluated per language before release. - citizen assistant · approved sources only - An agency's assistant may only say what is approved, and must answer to every citizen. - We build knowledge systems and assistants that cite their approved sources and route the rest to a caseworker, and the multilingual data that lets them serve every citizen. Civilian use cases only. - cited citizen and caseworker assistants · multilingual language data - [Public sector AI at YPAI](https://ypai.ai/solutions/public-sector/) - [Cited knowledge assistants](https://ypai.ai/enterprise-knowledge-assistants/) - [Multilingual speech data](https://ypai.ai/speech-data/) - answer highlighted in the regulation text · v2026-03 in force - sealed · sha-256 over question, passage bytes, version, approval time - out of scope · routed to a named caseworker #### One answer, sealed to its bytes #### Housing benefit regulation Applications are assessed against the regulation in force on the date of receipt. The following provision applies to late submissions. Applications received later than this are assessed only where statutory exceptions apply. - up to 14 days after the deadline - Late applications are accepted up to 7 days after the deadline where the delay is documented. - up to 7 days after the deadline #### can I still apply after the deadline #### can my debt be waived sealed under v2026-03 · source is superseded · answer withheld seal · sha-256(question ‖ passage bytes ‖ version ‖ approved time) seal broken · source changed · answer withdrawn ### Education - two points matched · one open · the educator decides - Tutoring and assessment assistants that match a learner's answer to the rubric and leave the decision with the educator. - Learning interactions, item pools and multilingual speech annotated by licensed educators and evaluated per subject. - assessment against a rubric · the educator decides - An assistant may support a grade, never issue one. Every point it awards traces to the learner's own words. - We build tutoring and assessment workflows that match an answer to the rubric and hand the decision to the educator, and the learning data they run on, reviewed by licensed educators and curriculum specialists, with the Article 10 evidence a high-risk classification asks for. - rubric-matched assessment support · educator-reviewed learning data - [AI for education at YPAI](https://ypai.ai/solutions/education/) - [Model and content evaluation](https://ypai.ai/data-solutions/evaluation/) - every awarded point traced to one sentence of the answer, with its score - the grade is proposed · the educator confirms, on the record #### The rubric · every point traced to the learner's own words, the educator decides Explain why a control group is needed in an experiment. - Isolates the effect of the treatment from other variables. - Provides a baseline to compare the treated group against. - Accounts for improvement caused by expectation alone, the placebo effect. A control group gives you a baseline to compare the treated group against. Without it you cannot tell whether the change came from the treatment or from something else that happened during the study. The study also needs enough participants to be meaningful. confirmed by the educator · on the record the educator may award or withhold any point run it on your own · minilm · in this tab load the model · about 22 MB ready · edit the answer and match it again matching every point to its closest sentence each rubric point is matched to the closest sentence by embedding similarity · a point below the threshold stays open · the educator confirms - [Media and platforms](https://ypai.ai/solutions/media-and-platforms/) ## Delivery designed for procurement and compliance review. Every engagement, whether it covers one service line or both, runs on Controlled Delivery: scope and acceptance criteria defined before work begins, rights and provenance documented, data and systems isolated, and human review, change control and delivery records inspectable throughout. Applicable GDPR, EU AI Act and EEA requirements are treated as controls the engagement is designed around and documented against. We run the delivery. You retain configuration authority. - QA sampling and review, run by YPAI. - Data, region and erasure stay yours. - Approval gates stay visible at every stage. Validation can begin with a customer-specific pilot scoped around the workflow, delivery requirements and agreed acceptance criteria. Commercial terms, setup costs and responsibility for non-accepted work are defined during scoping and written into the relevant SOW, acceptance terms, DPA and supporting documents. Customers review results in a controlled workspace before deciding whether to proceed to production. What a pilot demonstrates, it demonstrates against your requirements and acceptance criteria. ## Working outside these industries? - AI Data & Evaluation covers collection, annotation and evaluation for model development. - [Data and evaluation for AI companies](https://ypai.ai/solutions/ai-companies/) - AI Implementation covers assistants, agents and automated workflows. - [Enterprise AI services](https://ypai.ai/enterprise-services/) - Learning, assessment and language work for education providers. - [AI for education](https://ypai.ai/solutions/education/) - Rights-cleared dataset sourcing and licensing sits inside AI Data & Evaluation. - [Data services](https://ypai.ai/data-solutions/) Not sure where the workflow fits? Describe it below and YPAI will route it to the right delivery team. ## Two sentences set the route. The workflow you run, and the constraint it runs under. As you write, the route reads itself from your words. A named project lead confirms it and replies with feasibility. A named project lead reads the workflow and the constraint, and replies with feasibility and the route. YPAI will use the information provided to evaluate and respond to your scoping request. also on the same two lines - moderation and detection data · newsroom and content systems - in-cabin voice · perception data and evaluation - [Automotive and mobility](https://ypai.ai/automotive/) - clinical speech and imaging · EEA processing by default - [Healthcare](https://ypai.ai/solutions/healthcare/) - caseworker assistants · document AI for case intake - [Public sector](https://ypai.ai/solutions/public-sector/) - visual QC and inspection · capture from the plant - [Industrial and energy](https://ypai.ai/solutions/industrial-energy/) - document-heavy workflows · a decision you can audit - [Financial services](https://ypai.ai/financial-services-ai-solutions/) - recorded where the product runs · annotated to the frame - scoped around your workflow · production a separate decision - [Pilots](https://ypai.ai/pilots/) - recorded where the product runs · consent on every item - [Data collection](https://ypai.ai/data-collection/) ## You name the workflow and the constraint. We set the route, and prove it where it runs.