I design and scale production-grade RAG, MLOps, Kubernetes, and cloud platforms — turning infrastructure into intelligent, observable, and reliable systems.
I'm a Senior Platform & Cloud AI Engineer with 7+ years of experience building and scaling cloud infrastructure on AWS and Kubernetes. I've worked across fintech, logistics, and enterprise environments — shipping systems that handle millions of daily requests.
In 2024 I started going deep on AI engineering — not just using AI tools, but building production AI systems from scratch. My RAG Cloud Docs project demonstrates hybrid retrieval, RAGAS evaluation, and local LLM inference.
My edge is combining deep cloud infrastructure knowledge with hands-on AI engineering. I can ship a model and deploy it — most engineers can only do one of those.
Every architectural decision in my RAG system was validated with RAGAS scores before and after. I instrument, measure, and then decide.
I built the chunker from scratch because I wanted to understand the tradeoffs before tuning them. You can't debug a black box during an incident.
Most AI engineers build models that are hard to run in production. I start from containerized, observable, and cost-aware infrastructure.
The system that stays up at 3am beats the clever one that occasionally fails. I favor patterns that are simple, proven, and operable.
Running Ollama locally keeps sensitive runbooks and infrastructure context inside the network unless there is explicit justification.
Vector search finds meaning. BM25 catches exact terms. The best systems combine approaches intelligently instead of forcing one pattern.
A strong cloud and platform foundation with hands-on AI engineering, MLOps, observability, and infrastructure automation.
Production-style systems built from scratch — with real engineering decisions, evaluation scores, and infrastructure thinking.
Production-grade Retrieval-Augmented Generation system for querying cloud infrastructure documentation. Hybrid search, RRF fusion, local Ollama inference, FastAPI backend, and Streamlit chat UI.
Recursive chunking, hybrid retrieval, RAGAS evaluation across 28 Q&A pairs, and local execution with zero API cost during development.
End-to-end MLOps pipeline for infrastructure incident severity classification with automated training, evaluation gates, and KS drift detection.
Agentic system using LangGraph and MCP to automate cloud operations workflows such as incident response, cost analysis, and infrastructure review.
A production-grade RAG system built from scratch. No framework magic — every layer understood and owned.
On-call engineers spent 20+ minutes searching through runbooks during incidents. Keyword search returned too many irrelevant results.
Build a RAG system that answers in natural language, cites its source, and runs locally with Ollama so data stays inside the network.
Sub-second query latency, 0.929 faithfulness, 1.000 context recall, and 0.945 context precision across domain Q&A pairs.
Vector search understands meaning, while BM25 catches exact flags, error codes, and version numbers. RRF combines both.
Chunk size directly affects answer quality. 512 tokens with 50-token overlap gave the best balance of context and precision.
Zero API cost, local inference, and no sensitive infrastructure documents leaving the network.
Evaluation made every tuning decision measurable instead of subjective.
7+ years across fintech, logistics, and enterprise — shipping infrastructure that scales.
Industry certifications validating cloud architecture and infrastructure expertise.
Open to senior Cloud AI Engineering, MLOps, Platform Engineering, SRE, and Infrastructure Engineering roles.