appendix

Everything else

The supporting detail — tooling I actually reach for, smaller things I've built, and coursework. The research lives on the main page.

tooling

ML & NLP from scratch — linear/logistic/softmax regression, KNN, k-means, MLP, bag-of-words, TF-IDF, CBOW, skip-gram, byte-pair encoding, naïve Bayes and an LSTM sentiment classifier, all in pure NumPy/PyTorch with no black boxes. github ↗
Abler — real-time multimodal accessibility assistant for the visually impaired: object detection, depth estimation, speech and gesture recognition. Won eXabyte; Best Design at Hacknoware. github ↗ demo ↗
IPL 2025 predictor — decision trees, random forests, MLP classifiers and Monte Carlo simulation over a full EDA pipeline, plus a score-based statistical query system. Won Brain Dead 2025. github ↗
Vidvaan — research-paper aggregator unifying DBLP, arXiv, OpenLibrary and OpenAlex, with Gemini summarisation.
Thomas Raw — agentic travel planner on Gemini + LangChain; plans itineraries, budgets, and identifies locations from photos. github ↗
MentL — mental wellness assistant; Best Design Award, Hacknoware 2025.
Where Am I? — browser-based geography guessing game.
Model frontends — PPG encoder and SMS spam classifier interfaces built to make the underlying models usable; awarded at Tensor Quest, Instruo 2025.

deeplearning.ai · google cloud

Deep Learning Specialization Convolutional Neural Networks Sequence Models Structuring ML Projects Improving Deep Neural Networks GenAI with Gemini Vertex AI Studio

pre-university

All India Rank 6, West Bengal Rank 5 — CISCE Class 10 board examinations 98.8% aggregate · M. P. Birla Foundation HS School, Kolkata
ISC 95.3% / ICSE 98.3% 2010 — 2024

Global Rank 12 — International Mathematical Championship 2022
1st — national Essay-cum-Elocution contest, Indian Institute of Metals 2023 · paper on the role of AI/ML in materials science