GenAI Engineer

Job Title: GenAI Engineer (agents / RAG / evals)

Location: Hyderabad (Onsite)

Senior • 5-8+ Years Experience

About the Role

We're looking for a GenAI Engineer to build the agents themselves the prompts, tools, memory, retrieval and guardrails and tune them to real accuracy thresholds in production.

What You'll Do

• Build platform agents: Supervisor/Router, Retrieval, Policy Guard, Evidence & Audit, Notification, Human-Escalation.
• Build domain agents for the target process: Role Identifier, Approver Finder, Conflict/SoD Checker, Eligibility-Gate, Orchestrator, Deactivation/Leaver-Mover.
• Model Veeva Vault's access-control concepts security profiles, roles, groups, and Dynamic Access Control (DAC) into the knowledge graph so agents reason correctly about who should get access to what.
• Translate Veeva Vault integration and entitlement data (users, security profiles, DAC rules, product/domain access) into agent-ready context via retrieval and tool calls.
• Design and build the RAG pipeline: ingestion, chunking, embeddings, re-rank, citation and grounding.
• Build the knowledge graph (e.g. on neo4j) covering role/entity relationships and the vector-store schema.
• Own prompt architecture, the system-prompt library and its versioning; configure Bedrock guardrails.
• Seed golden datasets with subject-matter experts and run tuning cycles against evaluation findings.
• Optimise latency and cost model routing, caching, prompt compression.
• Build explainability for reviewers rationale, citations, confidence and drift instrumentation.

What You'll Own

• 12+ working agents
• Prompt library
• RAG pipeline
• Knowledge graph model
• Guardrail configuration
• Model cards & knowledge-base lineage
• Agent runbook and tuning playbook

What You'll Bring

• 4-7+ years building LLM/GenAI applications, including 1-2+ years building multi-agent systems in production.
• Hands-on experience with AWS Bedrock (or equivalent) and an agent-orchestration framework (AgentCore, LangGraph, or similar).
• Strong RAG engineering skills: chunking strategy, embeddings, re-ranking, grounding and citation.
• Experience modelling and querying knowledge graphs (neo4j preferred) and vector stores.
• Strong Python skills; prompt engineering and prompt-library/version-management discipline.
• Working knowledge of Veeva Vault's security/entitlement model security profiles, roles, groups, and Dynamic Access Control (DAC) setup and how to ground agent decisions in it.
• Veeva integration exposure (Vault REST API / VQL) for pulling user, role and access data into RAG/knowledge-graph pipelines is a strong plus.
• Experience running structured tuning cycles against evaluation feedback in partnership with subject-matter experts.
• Working knowledge of LLM guardrails, cost/latency optimization (model routing, caching, prompt compression).
• Comfort explaining model behavior (rationale, citations, confidence) to non-technical business reviewers.