Hey There! I am Snehashish Laskar.

Feb 2026 – present · nebula-assist.com

Production AI assistant platform. Containerised with Docker, fronted by Nginx. Bidirectional Google Calendar sync with conflict-free scheduling for focus sessions, and CI/CD pipelines that build, test and deploy web and mobile clients.

github.com/snehashish090/GradieX

A neural network library written from scratch in C++. Forward and back propagation, gradient descent and the tensor maths underneath — no external ML frameworks. Modular enough to stack arbitrary layers and swap activation and loss functions.

github.com/snehashish090/prodbot

Documentation agent for small teams. Runs as a git post-commit hook, reads the diff, writes changelogs, feature docs and architecture diagrams. Routing is plain Python rather than model calls, so a typical commit costs two API calls. Output is staged, never auto-committed — a human still reviews it.

github.com/snehashish090/connect4

Connect 4 agent trained with tabular Q-learning over 500,000 episodes and epsilon-greedy exploration. Q-table serialised to JSON so training resumes across sessions. Board logic and win detection hand-rolled; plays a human in the terminal.

github.com/snehashish090/ChopSticksAI

Reinforcement learning agent for the Chopsticks hand game. Full rules engine, move generation and a Q-learning decision layer that plays consistently strong games.

github.com/snehashish090/GenreDetect

End-to-end classification pipeline that infers music genre from structured audio features. Preprocessing, feature engineering and supervised training with metric-driven model selection. Packaged so retuning or swapping the dataset is a config change.