I am a Deep learning & Machine learning Engineer currently in my 6th semester at Academy of Technology, West Bengal. Beyond academics, I am a recognized Kaggle Expert — a competitive title earned through notebook medals and strong finishes on real-world datasets, not just coursework.

My work covers the complete ML pipeline: from exploratory data analysis and feature engineering, to training and fine-tuning production-grade models, to deploying them as real-time APIs that real users can interact with. I have built and deployed six live projects spanning conversational AI, medical image classification, fraud detection, churn prediction, rent forecasting, and NLP-based recommendation — each with a working public URL, not just a GitHub link.

My most recent work involves fine-tuning large language models: I fine-tuned facebook/blenderbot-400M-distill on the blended_skill_talk dataset, achieving a validation perplexity of 14.16, and deployed it as a containerized FastAPI service on HuggingFace Spaces with a full CI/CD pipeline via GitHub Actions. I also authored DataDiagnose, an open-source Python library published on PyPI with zero external dependencies that automatically audits ML datasets for data quality issues before training begins.