Technical Lead

End-to-End GenAI Solutioning
Architect and build production-grade GenAI & Agentic AI applications using Python with LangGraph for multi‑agent stateful workflows and LangChain for tool use, retrieval, and orchestration.
Model Selection & Configuration
Evaluate and configure foundation models (OpenAI, Anthropic, Azure OpenAI, Llama, Mistral, Bedrock models) including prompt engineering, system prompting, function/tool calling, temperature/top‑p tuning, context window optimization, and guardrails.
Retrieval-Augmented Generation (RAG)
Design high‑quality RAG pipelines: chunking strategies, embeddings (OpenAI, Cohere, BGE), vector stores (FAISS, Amazon OpenSearch, Pinecone), query rewriting, reranking, and evaluation of retrieval quality.
Agentic & Workflow Orchestration
Implement LangGraph graphs (nodes/edges, message/state management, checkpoints), multi-step tools, human‑in‑the‑loop, and error/retry policies; build LangChain agents (ReAct, Tool‑former styles) with reliable I/O schemas.
Data & Evaluation
Build datasets for prompts/test cases; implement offline evals (faithfulness, answer relevance, toxicity), online A/B tests, and observability (traces, token usage, latency) with tools like MLflow/Weights & Biases/OpenTelemetry/LangSmith.
Secure & Scalable AWS Implementation
Ship on AWS using Lambda / API Gateway / ECS or EKS, S3, Bedrock, SageMaker endpoints, DynamoDB/RDS, SQS/SNS, CloudWatch/X-Ray, KMS, Secrets Manager, and implement VPC, IAM least privilege, and cost controls.
Backend Engineering Excellence
Develop FastAPI/Flask microservices for inference, RAG, and async jobs; implement pagination, idempotency, caching, rate limiting, and typed contracts (pydantic) with robust unit/integration tests and contract tests.
DevOps, MLOps & CI/CD
Build IaC (Terraform/CDK), containerize with Docker, manage CI/CD (GitHub Actions/GitLab/Jenkins), automate model/package rollouts, blue‑green/canary deploys, and set up feature flags and rollbacks for model/prompt versions.
Compliance, Safety & Governance
Enforce PII/PHI controls, data retention, encryption in transit/at rest, consent & audit logging; implement safety filters, content moderation, jailbreak defenses, and document Model/Pipeline Cards and Data Lineage.
Collaboration & Leadership
Partner with product, data, and security; review designs/PRs, mentor engineers, manage technical debt, create best practices/playbooks, and drive measurable outcomes (latency, accuracy, cost per request, adoption)."