Datazymes Analytics -Lead AI Engineer (Agentic Systems)

WHAT WILL YOU DO: 1. Architecture & Engineering • Design and implement complex components of agentic pipelines multi-agent graphs, tool orchestration layers, retrieval modules, and memory systems — using LangGraph, AutoGen, CrewAI, or equivalent. • Take ownership of full sub-system designs: define agent topology, data flows, API contracts, and failure handling for a bounded scope, escalating trade-offs to the Sr. Architect. • Build and optimise production RAG pipelines: document ingestion, chunking strategy, embedding selection, hybrid search, retrieval evaluation, and latency tuning. • Integrate agentic systems with pharma data platforms (IQVIA, Symphony, Komodo, Veeva) via REST, event-driven hooks, and batch pipeline patterns. • Own observability for your components: instrument trace logging, cost metrics, drift alerts, and evaluation harnesses using LangSmith, Helicone, or equivalent. • Lead CI/CD for owned modules: containerisation (Docker/Kubernetes), automated test suites, staging gate criteria, and rollback procedures. 2. Pharma Domain Application • Translate medical affairs, commercial analytics, and clinical ops requirements into agent component specifications with minimal supervision. • Apply 21 CFR Part 11 auditability, HIPAA-compatible data handling, and GxP traceability patterns to every component you build. • Build intelligent document processing pipelines for pharma content: drug labels, clinical study reports, HEOR dossiers, and regulatory submissions. • Contribute to KOL mapping, competitive intelligence, and signal detection agents with domain-aware retrieval and reasoning strategies 3. Technical Leadership & Mentorship • Serve as the day-to-day technical reference for AI Engineers on your pod: code review, design feedback, unblocking implementation issues. • Lead component-level design reviews; surface architecture risks to the Sr. Architect or Associate Director before they reach staging. • Pair with junior engineers on hard problems; document patterns and decisions in the team’s shared knowledge base. • Represent engineering quality in client-facing technical discussions; translate complex trade-offs into plain language. • Contribute reference implementations and guardrail templates to the firm’s internal agentic AI playbook WHAT YOU BRING 1. Technical depth (Required) • 5–7 years in software or ML engineering; 2+ years building and shipping production LLM or agentic AI systems. • Hands-on proficiency with at least two agentic frameworks (LangGraph, LangChain, AutoGen, CrewAI); you have debugged framework internals, not just followed tutorials. • Direct SDK experience: Anthropic Claude API (tool use, streaming), OpenAI Assistants API, or Vertex AI Agent Builder. • Python mastery: production-quality code, type annotations, unit and integration tests, packaging, and performance profiling. • RAG pipeline depth: embedding model selection, vector stores (Pinecone, Weaviate, pgvector), hybrid retrieval, RAGAS or custom evaluation harnesses. • Cloud deployment: AWS, Azure, or GCP. Docker, Kubernetes, IaC basics (Terraform or CDK), CI/CD pipelines. • Agent observability: LangSmith, Helicone, or equivalent — you have diagnosed latency, cost, and quality issues in production traces. 2. Pharma / Life Sciences Domain (Required) • Working knowledge of pharma commercial data: Rx/claims, NPI-level analytics, brand performance metrics. • Experience operating in regulated data environments: GxP, 21 CFR Part 11, HIPAA compliant data handling. • Exposure to at least one of: medical affairs analytics, RWE, clinical operations data, HEOR/market access, or regulatory intelligence. 3. Leadership & Communication (Required) • Track record of shipping 2+ agentic or ML systems to production — not just proof-of-concepts — with documented performance benchmarks. • Experience acting as technical lead or senior reviewer for at least one junior engineer or cross functional delivery workstream. • Ability to write crisp component specifications and communicate architectural trade-offs to both engineers and non-technical stakeholders. 4. Good to Have • MCP (Model Context Protocol) implementation experience. • Experience with Veeva Vault, Medidata, IQVIA, or Symphony Health platform integrations. • Familiarity with knowledge graphs (Neo4j, Amazon Neptune) for pharma entity modelling. • Exposure to RLHF, fine-tuning, or model adaptation workflows. • Prior consulting or services-firm delivery experience