Software engineer with five years building AI/ML systems in production. I work across the full stack - LLMs, classical ML, deep learning, and the infrastructure underneath.
Generative AI & LLMs RAG pipelines, agentic systems, multi-agent routing, prompt engineering. Production work with GPT, Claude, and open-source foundation models.
Classical ML & Deep Learning Classification, clustering (HDBSCAN, LDA), topic modeling, BERT and transformer fine-tuning, XGBoost. Applied to real problems with messy data.
ML Infrastructure Model serving, GPU inference, vector databases, Kubernetes autoscaling, Redis caching, latency optimization.
Backend & Distributed Systems FastAPI, Kafka, PostgreSQL, event-driven microservices. Backend design as part of the AI system, not separate from it.
Agentic routing Intent classification → dynamic agent dispatch
Fallback chains, confidence thresholds, retry logic
RAG systems Chunking strategies, dense + sparse retrieval
Reranking, context compression, hallucination mitigation
Document intelligence Schema extraction, field validation, structured output
Azure AI Search, vector + keyword hybrid retrieval
Log & event analytics Temporal windowing, density-based clustering (HDBSCAN)
Noise filtering, pattern surfacing at scale
ML serving GPU inference on Kubernetes, horizontal autoscaling
Redis-backed caching, p99 latency budgeting
NLP pipelines BERT fine-tuning, topic modeling (LDA), similarity scoring
Batch inference, compliance classification
Languages Python · Java · SQL
AI / ML LangChain · HuggingFace · PyTorch · TensorFlow
FAISS · XGBoost · Transformers · OpenAI · Anthropic
Cloud Azure (AKS, AI Search, Cosmos DB)
AWS (SageMaker, EKS, Lambda, S3)
Infra Kubernetes · Docker · Redis · Kafka
Backend FastAPI · Flask · PostgreSQL · Node.js · Microservices
| Project | What it is |
|---|---|
| PromptIQ | Prompt evaluation engine - TF-IDF task classification, model-fit scoring, no LLM in the eval loop |
| self-healing-rag-pipeline | Minimal RAG pipeline built to understand the retrieval layer properly, not abstract it |
| agent-routing-benchmark | Intent-based multi-agent router with a benchmark comparing routing strategies |
| llm-eval-kit | Lightweight harness scoring LLM outputs on faithfulness, relevance, and coherence |
Building open-source LLM eval and prompt quality tooling. Working through system design and DSA alongside a full-time job.
Open to conversations about AI/ML engineering roles.