Sagnik Roy

Distinguishing the signal from the noise

+ +

About Me

Deep Learning & MLOps Developer & NLP Specialist

I am driven by the power of data and advanced computation to create intelligent systems. My passion lies in building, training, and deploying robust Deep Learning models that tackle complex, real-world challenges.

Specializing in Natural Language Processing (NLP), I am proficient in leveraging GRUs/LSTMs & transformer architectures for tasks like text generation, sentiment analysis, and semantic search.

Python PyTorch NLP Transformers BERT Models FastAPI Docker

Check my projects out

GRU/LSTM based sentence AutoCompletion Model

Created two GRU & LSTM based machine learning model from scratch for autocompleting sentences by character level predictions using NLP. Compared performance & deployed it via FastAPI & Docker.

PyTorch Python Natural Language Processing Docker FastAPI

View Code Live Demo

TRUST NSFW Detection soft (w/ near SOTA benchmarks)

TRUST-NSFW-Classifier is a high-performance, deep-learning solution engineered to solve the "Binary Bottleneck" of modern content moderation. While traditional classifiers often fail at the nuance between "artistic" and "explicit," TRUST leverages state-of-the-art Vision Transformer (ViT) architecture to deliver granular, multi-dimensional safety scores.

PyTorch Python Natural Language Processing Docker FastAPI

View Code Live Demo

Notes recognizer & memory based chat System

This project is a subpart of a bigger personal project called Mynd. This system recognizes handwriting from seemingly infinite spaces and makes a intelligent query based chat systems with memory

Retrieval Augmented Generation Vector DB LLMs(Gemini) MLKit System Architecture

Details

+ +

Message me on