I'm an AI undergraduate with a full-stack (MERN) engineering background, currently deepening my focus on machine learning and deep learning — and now moving into agentic AI systems: LLM-powered agents, tool use, and multi-step reasoning pipelines.
- 🎓 Studying Artificial Intelligence
- 🧠 Solid foundation in Machine Learning, Deep Learning & Neural Networks
- 🤖 Currently exploring Agentic AI — building LLM agents that can reason, plan, and use tools
- 🌐 Full-stack developer (MERN) — I build the systems that ship ML models, not just the models themselves
- 📚 Continuously learning: currently going deeper into LLM architectures, fine-tuning, and agent frameworks
AI / ML
Python · PyTorch · HuggingFace Transformers · scikit-learn · SciPy · NumPy · Pandas
Frontend
React · Tailwind CSS · JavaScript
Backend
Node.js · Express.js · NestJS
Database
MongoDB · MySQL
Machine Learning / Deep Learning
Python ▓▓▓▓▓▓▓▓▓░ 85%
PyTorch ▓▓▓▓▓▓▓▓░░ 80%
Transformers ▓▓▓▓▓▓▓░░░ 75%
scikit-learn ▓▓▓▓▓▓▓▓░░ 80%
SciPy ▓▓▓▓▓▓▓░░░ 70%
Frontend
React ▓▓▓▓▓▓▓▓░░ 80%
JavaScript ▓▓▓▓▓▓▓▓▓░ 85%
Tailwind CSS ▓▓▓▓▓▓▓░░░ 70%
Backend
Node.js ▓▓▓▓▓▓▓▓░░ 80%
Express.js ▓▓▓▓▓▓▓▓░░ 80%
NestJS ▓▓▓▓▓▓░░░░ 60%
Database
MongoDB ▓▓▓▓▓▓▓▓░░ 80%
MySQL ▓▓▓▓▓▓░░░░ 60%
Fine-tuned a T5-base transformer end-to-end to translate English questions into SQL queries, trained on the Spider dataset across 100+ database schemas.
- Built a schema-aware input pipeline, full training/validation loop, and checkpointing
- Achieved 97.5%+ validation token accuracy after fine-tuning
- Went beyond surface metrics with exact-match evaluation to catch real query-correctness issues (e.g. GROUP BY column handling, table disambiguation) that token accuracy alone missed
Stack: Python · PyTorch · HuggingFace Transformers · T5