ML Engineer (who builds stuff that actually works)
- Build and deploy ML models that make sense (and predictions)
- Fine-tune LLMs using LoRA, PEFT — the good stuff, not the buzzwords
- RAG pipelines with LangChain + vector DBs that actually retrieve relevant info
- Wrap it all up with FastAPI, Docker, and a touch of ✨ sanity ✨
-
Started training transformers before understanding attention
→ Drowned in tensor hell. Came back knowing exactly how and why attention works. -
Overengineered LSTMs with too many layers
→ Thought “deeper = better.” Learned that simplicity + the right loss function > 50 layers of pain. -
Assumed FAISS “just works”
→ Spoiler: it doesn’t. Learned to tune it, index properly, and actually measure retrieval quality. -
Wrote messy APIs during my first FastAPI project
→ Came back with better routing, modular structure, and swagger docs that don’t scream "help me."
- Transformers from scratch — NumPy to PyTorch (yes, it actually worked)
- LSTM models for real-world problems (SRU optimization, aquaponics monitoring)
- RAG systems + LLM fine-tuning in freelance projects (and nobody cried)
- Currently building on Jetson Orin Nano for real-time scrap detection (fancy, I know)
PyTorch • FastAPI • scikit-learn • NumPy • Pandas • LangChain• Azure
I don’t just read AI papers — I turn them into working code (and clean up after the explosion).