I build production AI systems — from multi-agent architectures and LLM fine-tuning to classical ML pipelines. Full lifecycle: research, training, deployment, monitoring.

• Multi-Agent Systems — Built a legal analysis system automating contract review: 30 min → 2 min, 95% recall on critical risk clauses
• Production AI Pipelines — Engineered a GenAI marketing system for a major enterprise: serves ~1,000 users, +36% client engagement
• Agentic Workflows — Designed a customer-support agent serving 3,000+ monthly users, reducing operator workload by 2x
• Predictive ML — XGBoost models for equipment failure prediction (+23% reliability) and customer behavior forecasting (+18% retention)

Core: Python, PyTorch, Transformers, LangChain, LangGraph, FastAPI LLM/NLP: Unsloth, spaCy, Elasticsearch, ChromaDB, Prompt Engineering ML: Scikit-Learn, XGBoost, CatBoost, PySpark, NumPy, Pandas Infra: Docker, Airflow, MLflow, RabbitMQ, AWS, PostgreSQL Languages: C++, SQL, R

Google DeepMind Hackathon — Built "AI-Psychologist" with Gemini 3 Pro: real-time CBT reasoning + voice + vision analysis. Fully functional MVP in under 6 hours.

Kaggle — 10+ competitions, top 3% finishes. Fine-tuned Qwen-2.5 for mathematical misconception detection.

Education — IBM RAG & Agentic AI Certificate | DeepLearning.AI Deep Learning Specialization | Stanford ML Specialization