Chiri is an AI managed service provider for company-specific AI.
The Role
Build and deploy a secure, offline document-intelligence system for RFQ screening. Responsibilities include PDF and drawing processing, OCR, multimodal model deployment, structured citation-grounded extraction, validation rules, audit logging, evaluation frameworks, and local versus Azure OpenAI comparisons. The role also includes packaging signed offline releases, documenting deployments, supporting onsite stakeholder sessions, and assisting client IT teams within strict CUI and air-gapped security boundaries.
Summary Generated by Built In
About Chiri
Chiri collapses fragmented software into one governed operating layer for mid-sized businesses: companies with valuable workflows and real urgency, but no internal AI organization to take a use case to production.We combine our platform with forward-deployed expertise to connect customers' existing systems, build what they need, and operate it in production.Increasingly, buyers are asking ChatGPT, Perplexity, Gemini, Claude, and other AI systems what they should use. Chiri must be one of their top answers.
Role Overview
Key Responsibilities
Required Skills and Experience
Preferred Qualifications
What We Offer
Security & Data Boundaries
Our SCOUT Values
Our values are not wall art. They describe how we expect people to operate.
Seek to Understand
Call It Out
Ownership Mentality
Use Human Judgment
Talk with Precision
Role Overview
We're looking for an Applied AI Engineer to design, build, and deploy an AI-powered document intelligence solution for a Silicon Valley electronics manufacturer operating in the defense supply chain.
This is a hands-on engineering role focused on building a production-ready, zero-egress RFQ (Request for Quote) screening service that processes complex technical documents entirely within the client's security boundary.
The solution will help the client's estimating team review incoming RFQ packages, identify potential disqualifiers, extract quality requirements, analyze material specifications, and flag missing or conflicting information.
The service will process approximately 70 RFQ packages per day, each containing 3–30 pages of engineering drawings and 10–20 quality clauses. Today, reviewing a single package can take an estimator 1.5–3 hours.
You'll build an end-to-end document processing pipeline using Python, open-weight multimodal models, and modern AI engineering techniques. The system must produce structured, citation-grounded outputs, with every extracted value traceable to its source document and page.
This role requires strong technical ownership, practical experience deploying AI models locally, and a disciplined approach to security and data handling. You'll work directly with Chiri's engineering team and the client's estimating, quality, and IT stakeholders to deliver a solution that supports human decision-making.
The system supports human decisions. It never makes a bid or decline decision itself.
- Design, develop, and maintain the end-to-end RFQ screening pipeline, from PDF ingestion and document classification to AI-powered extraction, rule application, and structured report generation.
- Build document-processing workflows that extract text from PDFs, render drawing pages as images, classify document types, and process complex technical content.
- Deploy and optimize open-weight multimodal models for fully offline inference on a single GPU server.
- Select appropriate models and quantization strategies to operate within approximately 48 GB of GPU memory.
- Develop and refine prompts and extraction logic for quality clauses, material tables, engineering drawings, and technical specifications.
- Implement citation-grounded extraction, ensuring every value includes a page reference and supporting source text.
- Build validation and rule-based logic to identify potential disqualifiers, compare extracted materials against stock lists, and flag missing or conflicting values.
- Develop the same Chiri application layer against Azure OpenAI through Azure Government and a private endpoint, enabling a clean comparison between local inference and the Azure-based option.
- Implement a local data store and audit logging system using SQLite or SQL Server Express to record model calls, input file hashes, and generated reports.
- Package application releases as signed, offline-installable artifacts with pinned dependencies, a Software Bill of Materials (SBOM), release hashes, and supporting documentation.
- Produce data-flow diagrams and technical documentation for each deployment option to support the client's System Security Plan.
- Build an evaluation harness that allows the client's estimating team to score report accuracy by field.
- Improve extraction quality using aggregate evaluation metrics and scorecards without accessing or receiving the client's underlying production data.
- Lead or support the onsite Week 1 working session with estimating, quality, and IT stakeholders to define accuracy targets, select sample RFQs, and establish the disqualifier list.
- Support the client's IT team with installation, configuration, and troubleshooting within the agreed security and access boundaries.
- Collaborate with Chiri engineers and client stakeholders to translate business requirements into practical, secure, and maintainable technical solutions.
- Deliver a final accuracy scorecard comparing the local and Azure-based options, broken down by field.
- Contribute technical recommendations and implementation estimates for the potential Phase 1 engagement.
Required Skills and Experience
- 5+ years of professional software engineering experience, with a strong track record of shipping production systems.
- Strong professional experience with Python, including backend development, data processing, and AI-powered applications.
- Hands-on experience building LLM or vision-language model (VLM) extraction pipelines for complex documents, including PDFs, scanned pages, tables, and technical drawings.
- Experience implementing structured AI outputs, source grounding, and citation-based extraction.
- Hands-on experience self-hosting open-weight models on GPU hardware, using tools such as vLLM, llama.cpp, Ollama, or Hugging Face TGI.
- Experience selecting and optimizing models and quantization levels for local inference, including working within GPU memory constraints.
- Experience building document-processing pipelines that include PDF text extraction, page rendering, OCR, and document classification.
- Experience deploying AI systems in air-gapped or offline environments, including offline installation, pinned dependencies, and no runtime internet access.
- Experience with Azure OpenAI and Azure networking, including private endpoints and keyless or customer-managed authentication.
- Strong understanding of software supply-chain security, including signed releases, SBOM generation, and reproducible builds with published hashes.
- Experience building evaluation frameworks for AI extraction systems, including per-field accuracy metrics and systematic quality improvement.
- Solid SQL skills and experience designing local data stores and audit logs using SQLite, SQL Server, or comparable databases.
- Strong understanding of software architecture, system design, and engineering best practices.
- Experience owning technical deliverables from design through implementation, deployment, and ongoing support.
- Comfortable working onsite with non-technical stakeholders, including estimating, quality, and IT teams.
- Ability to explain AI behavior, limitations, accuracy, and technical tradeoffs in clear, practical language.
- Strong analytical and problem-solving skills.
- Excellent written and verbal communication skills in English.
- A disciplined approach to data security and the ability to work effectively without direct access to production data.
- Working knowledge of NIST SP 800-171, CMMC, DFARS 252.204-7012, CUI handling, and System Security Plans.
- Experience delivering AI solutions in defense, aerospace, electronics manufacturing, or other regulated environments.
- Familiarity with Azure Government, including service and model availability, quotas, private networking, and abuse-monitoring exemption processes.
- Experience working with printed circuit board (PCB) manufacturing specifications, including IPC standards, laminate materials and Tg, copper weight, surface finishes, solder mask, and rigid-flex materials.
- Familiarity with electronics quoting, manufacturability analysis, CAM workflows, or tools such as Frontline Integr8tor.
- Experience interpreting engineering drawings, fabrication notes, quality clauses, or flow-down requirements in aerospace and defense purchase orders.
- Experience deploying applications on Windows Server or Linux virtual machines managed by a client's IT team.
- Experience writing installation guides, operational runbooks, and technical documentation for customer-managed environments.
- Familiarity with Microsoft Exchange, Outlook, and Access integrations.
- Experience building rules engines or spreadsheet-driven business rules that can be maintained by business users.
- Experience with agentic and AI-assisted development workflows.
- Experience in consulting, forward-deployed engineering, or client-facing technical delivery.
- Demonstrated ability to deliver measurable business outcomes through AI and automation.
- The opportunity to build a real-world AI solution for a complex, high-impact manufacturing use case.
- Meaningful technical ownership across AI model deployment, document intelligence, system architecture, and secure software delivery.
- The opportunity to work with cutting-edge open-weight multimodal models and local inference infrastructure.
- Direct collaboration with experienced engineers, company leadership, and client stakeholders.
- A focused, fast-paced engagement where engineering decisions have a tangible impact on business operations.
- The opportunity to help define how enterprise-ready AI can operate within strict security and compliance boundaries.
- Potential to extend the engagement into Phase 1, subject to project needs and mutual agreement.
The data boundary defines the project.
The service runs entirely inside the client's CUI boundary and makes no outbound connections.
- No telemetry.
- No automatic updates.
- No external license checks.
- No transfer of client RFQs, drawings, or extracted data to Chiri.
- No copying or storing client production data outside the approved environment.
Chiri delivers a signed installation package with an SBOM and releases hashes. The client's IT team installs and operates the solution.
Support takes place through client-controlled screen-sharing. The engineer does not log in to client systems, receive client data, or take control of the client's keyboard.
The engineer must be comfortable developing, evaluating, and improving the solution while respecting these strict boundaries.
Engagement Details- Initial duration: Approximately 6 weeks (Phase 0).
- Potential extension: Phase 1, subject to project requirements and agreement.
- U.S. person status. Must be eligible to work under applicable U.S. person requirements, sign a written attestation, and comply with DFARS 252.204-7012 and CUI handling requirements.
Our SCOUT Values
Assume you might be wrong. Clarity before action.
See something, say something. Say the hard thing.
"That's not my job" does not exist.
Judgment, context, and care come from us. AI is a tool.
Concise. Intentional.
Skills Required
- 5+ years of professional software engineering experience shipping production systems
- Strong professional experience with Python for backend development, data processing, and AI applications
- Experience building LLM or vision-language model extraction pipelines for PDFs, scanned pages, tables, and technical drawings
- Experience implementing structured AI outputs, source grounding, and citation-based extraction
- Experience self-hosting open-weight models on GPU hardware using tools such as vLLM, llama.cpp, Ollama, or Hugging Face TGI
- Experience selecting and optimizing models and quantization levels within GPU memory constraints
- Experience building document-processing pipelines with PDF extraction, page rendering, OCR, and document classification
- Experience deploying AI systems in air-gapped or offline environments with pinned dependencies and no runtime internet access
- Experience with Azure OpenAI and Azure networking, including private endpoints and keyless or customer-managed authentication
- Strong understanding of software supply-chain security, signed releases, SBOM generation, and reproducible builds with published hashes
- Experience building AI extraction evaluation frameworks with per-field accuracy metrics and systematic quality improvement
- Solid SQL skills and experience designing local data stores and audit logs using SQLite, SQL Server, or comparable databases
- Strong understanding of software architecture, system design, and engineering best practices
- Experience owning technical deliverables from design through implementation, deployment, and ongoing support
- Comfort working onsite with estimating, quality, and IT stakeholders
- Ability to explain AI behavior, limitations, accuracy, and technical tradeoffs clearly
- Strong analytical and problem-solving skills
- Excellent written and verbal communication skills in English
- Disciplined approach to data security and ability to work without direct access to production data
- Working knowledge of NIST SP 800-171, CMMC, DFARS 252.204-7012, CUI handling, and System Security Plans
- Experience delivering AI solutions in defense, aerospace, electronics manufacturing, or regulated environments
- Familiarity with Azure Government service and model availability, quotas, private networking, and abuse-monitoring exemptions
- Experience with PCB manufacturing specifications, IPC standards, laminate materials, copper weight, surface finishes, solder mask, and rigid-flex materials
- Familiarity with electronics quoting, manufacturability analysis, CAM workflows, or Frontline Integr8tor
- Experience interpreting engineering drawings, fabrication notes, quality clauses, or aerospace and defense purchase-order requirements
- Experience deploying applications on Windows Server or Linux virtual machines managed by client IT teams
- Experience writing installation guides, operational runbooks, and technical documentation for customer-managed environments
- Familiarity with Microsoft Exchange, Outlook, and Access integrations
- Experience building rules engines or spreadsheet-driven business rules maintained by business users
- Familiarity with agentic and AI-assisted development workflows
- Experience in consulting, forward-deployed engineering, or client-facing technical delivery
- Demonstrated ability to deliver measurable business outcomes through AI and automation
- Eligibility to work under applicable U.S. person requirements and sign a written attestation
- Compliance with DFARS 252.204-7012 and CUI handling requirements
Am I A Good Fit?
Get Personalized Job Insights.
Our AI-powered fit analysis compares your resume with a job listing so you know if your skills & experience align.
Success! Refresh the page to see how your skills align with this role.
The Company
What We Do
We turn business problems into governed applications, agents, and targeted software. Our forward-deployed pods combine senior operators and engineers. Each pod embeds with client teams, learns how work happens, and ships production systems within weeks. Chiri supports AI adoption, AI optimization, and AI software engineering. We assess workflows, design operating models, and deploy digital workers into existing tools. We also audit scattered AI systems, add governance, consolidate systems, and measure return on investment. Chiri Brain provides the shared foundation for every deployment. Its ontology preserves company knowledge and operating context. Sherpa connects current applications and provides broad AI model access. Glacier applies security, permissions, validation, and audit controls. Client data remains client-owned. Each Statement of Work defines the custom deliverables and related terms. Chiri supports work across go-to-market, engineering, finance, and people operations. Our systems work inside tools such as Slack, HubSpot, Gmail, Teams, Salesforce, Jira, and Zendesk. We build from a first production application through enterprise scale. Chiri Brain remains after the pod steps back. This model gives companies lasting AI capability without requiring a separate internal AI company.
Why Work With Us
We value curiosity, ownership, speed, and people who love solving complex problems with simple solutions. If you’re excited about building a meaningful company, working with highly talented people, and making a real impact from day one, we’d love to meet you.
.png){kind=logo}
