• Cross-modal image registration for autonomous navigation
• Evaluation-driven, reproducible ML
• Fine-tuning LLMs (LoRA / PEFT)
• Speech-to-text systems
• RAG & retrieval pipelines

Voice in, answers out. Speech → retrieval → local LLM, with memory that actually persists.

Whisper LangGraph FAISS Ollama FastAPI Redis

Ask questions over financial docs. Fine-tuned for the domain; evaluated rigorously.

Flan-T5-XL ChromaDB HuggingFace

Matching what a descending spacecraft sees to a reference map of the Moon. No GPS. No margin for error.

Python OpenCV rasterio LOLA / TMC-2

Making speech recognition work for people who don't speak English. Iterated on error analysis, not just loss curves.

Faster-Whisper PyTorch

Languages — Python · C / C++ · SQL

ML / DL — PyTorch · TensorFlow · HuggingFace Transformers · LoRA / PEFT

Speech — Whisper · Faster-Whisper

LLM & retrieval — LangChain · LangGraph · FAISS · ChromaDB · RAG

Infra — FastAPI · Docker · Redis · Linux · Git

• Machine Learning Specialization — Coursera
• Multi-Agent Systems (crewAI) — DeepLearning.AI
• TensorFlow for AI, ML & DL — Coursera