Full-Stack AI & Access-Governance Engineer

Role: 

• • JP35478 will be more AI & security focused. 
    • Should have experience integrating LLMs with AI would be sufficient. 
  • JP35477 should have RAG experience. Working on natural language applications (like chatbots). 
  • JP35476 will be working on building agent orchestration application. 
  • All roles:
    • Roughly 5-10 projects (deliverables) recently given to this group to tackle before EOY.
• Required: 
  • 3+ years of experience required. 
  • Active Directory required. 
• Preferred: 
  • Pharmaceutical experience is HEAVILY preferred. 
  • Docker experience is a huge plus. 
• Education: 
  • Bachelor's degree (must be technical) required. 

The role

Build four AI tools for pharma BD/S&E analysts into supportable production services, and build the Deal Access Control Center — an Entra ID / Active Directory reconciliation service that governs access to all of them. Three interchangeable full-stack engineers; you move across backend, frontend, the LLM/agent layer, the Microsoft identity stack, and deployment.

What you'll build

• AI tools (BD Deals Engine, Due Diligence, Triage, Landing Page): add test coverage, move secrets to a vault + service accounts, fix CI/CD and dependency pinning, add observability, extend document extraction (PDF/PPT/XLSX/DOCX), enforce an outbound-data guard, and preserve the RAG + citation-verification quality.
• Deal Access Control Center: an idempotent, hourly reconciliation service that reads the deal "tent list", validates + gates entries (CDA, training), provisions deal-scoped Entra/AD security groups (APP-<Deal>-Owner/Editor/Viewer), and writes an immutable audit log. Integrate consumers by push (SharePoint, Teams, VDRs via Microsoft Graph) and pull (AI tools, Dataverse, ARCH at query time).

Technical requirements – Must have:

• Languages     
  • Python 3.11+, TypeScript, SQL
• Backend          
  • FastAPI (async), server-sent events / streaming
• Frontend         
  • React 18, Vite, Node
• AI / LLM            
  • LLM app development: agent tool-use loops, prompt engineering, streaming; RAG — embeddings, vector/similarity search, chunking, source grounding
• Microsoft identity      
  • Entra ID / Azure AD, Microsoft Graph API, security-group provisioning, SharePoint/Teams automation; RBAC, least-privilege, default-deny
• Infra / DevOps             
  • Docker, Kubernetes, CI/CD, secrets management; idempotent scheduled/reconciliation services
• Quality             
  • pytest + Vitest/Jest; adding tests to untested code
• Experience     
  • 3-5 years full-stack; comfortable owning an unfamiliar codebase end-to-end

Nice to have

• Identity governance (IGA — access certification, joiner-mover-leaver, audit logging) · Power Platform / Dataverse (row/column security) · knowledge graphs & Cypher · Claude Code extension points (skills, subagents, MCP connectors) · document parsing (PyMuPDF, python-pptx/docx, openpyxl) · VDR / external-party provisioning · regulated-enterprise deployment (proxy, SSO, internal gateways) · pharma BD / licensing / M&A domain.

Success

• Tools are tested, observable, and reproducibly deployed with managed secrets. Access is granted within one hour of a tent-list entry, zero entitlements live outside a deal security group, revocation is immediate, and audit evidence is produced on demand.