• MS in Data Science from George Washington University
• I build AI systems that move from research → production
• Experience across backend systems, ML engineering, and applied AI deployments
• Focused on scalable, explainable, and responsible AI solutions

Multi-step Agentic RAG workflow that converts due diligence documents into structured, citation-backed investment memos. Tech: Python · FastAPI · React · RAG · Pydantic

• Learned: Closed-book evidence gating with citation-backed outputs for high-stakes documents
• Learned: Human-in-the-loop review flows using confidence scoring and conflict detection.

Transformer-based NLP system for categorizing consumer complaints
Tech: PyTorch · BERT · FastAPI · React

• Learned: Fine-tuning transformer models for domain-specific text
• Learned: Handling imbalanced datasets using weighted loss + evaluation tuning

Agentic AI assistant for real-estate intelligence using natural language querying and retrieval-augmented reasoning.
Tech: LangGraph · PostgreSQL · FAISS · OpenAI

• Learned: Translating ambiguous user intent into executable database operations
• Learned: Designing guardrails for LLM-driven data access

• Retrieval quality matters more than model size in many RAG systems.
• Human review is often a feature, not a limitation.
• Explainability becomes essential in high-stakes domains.
• The hardest part of AI products is often the workflow around the model.