Nilabhra Chowdhury
Staff AI Engineer (Training) at Nscale
Dubai, Dubai, United Arab Emirates
Summary
👤
Senior🎓
Top SchoolNilabhra Chowdhury is a Staff AI Engineer with 11 years of experience building production-ready NLP/NLU systems that balance state-of-the-art performance with computational efficiency. He has led large-scale distributed training of LLMs (including Falcon variants across hundreds of GPUs), implemented low-level optimizations like CUDA-backed optimizers and Triton kernel updates, and operationalized models via Ray/Kubernetes pipelines. Nilabhra blends classical NLP and deep learning—often shipping hybrid solutions for noisy real-world data—and has contributed to the popular Flair NLP framework by developing and testing document embeddings and memory optimizations. Based in Dubai, he excels at translating research papers into robust code and seeks novel techniques that reduce compute while improving results, a throughline from early QA and retrieval work to recent LLM infrastructure leadership.
11 years of coding experience
8 years of employment as a software developer
Bachelor of Technology (B.Tech.) Computer Science and Engineering, Bachelor of Technology (B.Tech.) Computer Science and Engineering at West Bengal University of Technology, KolkataM.Sc Data Engineering, M.Sc Data Engineering at Jacobs University BremenCBSE Higher Secondary Science with Computer, CBSE Higher Secondary Science with Computer at B.D.M International
English
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (8)
pytorch10
word-embeddings10
machine-learning10
nlp10
python10
pytest9
testing9
sequence-labeling9
Programming languages (4)
Github contributions (5)
flairNLP/flair
Nov 2018 - Jan 2019
A very simple framework for state-of-the-art Natural Language Processing (NLP)
Role in this project:
ML Engineer
ML EngineerContributions:10 commits, 1 PR, 9 comments in 1 month
Contributions summary:Nilabhra primarily contributed to the development and testing of document embeddings within the Flair NLP framework. Their work involved adding a new `DocumentPoolEmbeddings` class, fixing typos, and implementing dynamic naming based on pooling modes. They also wrote tests for the new embeddings, ensuring proper functionality and integration. Furthermore, they made modifications to the sequence tagger model, adding a progress bar for verbose output, optimizing memory usage by clearing token embeddings after batch processing and implementing presorting of sequences.
nlppytorchnamed-entity-recognitionartsemantic-role-labeling
petuum/pirlib
Nov 2022 - Jan 2023
Contributions:18 reviews, 73 commits, 7 PRs in 2 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.