• Agentic AI Engineering — RAG pipelines (LangChain, vector search, hybrid retrieval + reranking), structured LLM output with schema validation and retry-on-failure, confidence-based grounding/hallucination filtering, content-safety guardrails, and agentic automation (custom Claude Code agents, MCP tooling, LLM-driven PR review and issue triage).
• Platform Engineering — Infrastructure-as-code with Terraform across multi-environment deployments, containerized services on Docker/Kubernetes/Azure Container Apps, event-driven workers (Azure Functions, Service Bus, Celery), CI/CD (GitHub Actions, Bitbucket Pipelines), and OpenTelemetry-based observability.
• Python Software Engineering — Async-first APIs with FastAPI and Django/DRF, strict typing (mypy), SQLAlchemy + Alembic migrations, test-driven backend design.
• AI Engineering — LLM integration across OpenAI, Azure OpenAI, and Claude; embeddings and vector search (Azure AI Search, MongoDB Atlas $vectorSearch, FAISS); token usage and cost accounting.
• Data Engineering — Relational + NoSQL data modeling (Postgres, MySQL, Cosmos DB, MongoDB), message-queue-driven ingestion pipelines, ETL for document/knowledge stores feeding retrieval systems.

• Building Tokenria — a local tool that breaks down LLM token spend by structural category (cache, tools, history, output) and separately tracks how much of what a model generates actually gets used, tagged by hand rather than auto-scored.
• Deepening production RAG patterns: retrieval grounding, hallucination confidence-filtering, and guardrail testing for LLM-backed platforms.
• Growing a Claude Code / MCP-based agentic tooling practice — custom agents for PR review, technical-debt triage, and dev workflow automation.

Open to collaborating on agentic AI systems, RAG infrastructure, and platform engineering for AI-native products.