Vedanuj Goswami
Research Engineer at Meta
Menlo Park, California, United States
Summary
🤩
Rockstar🎓
Top SchoolVedanuj Goswami is a Principal Research Engineer at Meta with a decade of experience specializing in deep learning, computer vision, NLP, and multimodal AI. He rose through engineering and research roles at Meta, contributing both production-quality backend systems and research tooling that bridge vision and language tasks. His open-source work on high-profile FAIR projects like vilbert-multi-task and mmf shows hands-on expertise in dataset integration, distributed training, dependency refactors, and robust test automation. Known for reducing engineering friction—e.g., removing unnecessary dependencies and hardening data loaders—he combines rigorous QA instincts with research-forward model development. Based in Menlo Park with an MS from Georgia Tech, he blends academic foundations with practical impact on large-scale multimodal research.
10 years of coding experience
Master of Science (MS), Computer Science, Master of Science (MS), Computer Science at Georgia Institute of TechnologyBachelor of Technology (B.Tech.), Computer Science, Bachelor of Technology (B.Tech.), Computer Science at National Institute of Technology Silchar
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (15)
data-set10
unit-testing10
computer-vision10
machine-learning10
pytorch10
python10
data-integration10
datasets10
modeling9
test-automation9
trainings9
software-quality9
deep-learning9
data-loading9
nlp8
Programming languages (4)
Github contributions (5)
facebookresearch/mmf
Jan 2019 - Sep 2021
A modular framework for vision & language multimodal research from Facebook AI Research (FAIR)
Role in this project:
Back-end Developer & QA Engineer / Test Automation Engineer
Back-end Developer & QA Engineer / Test Automation EngineerContributions:165 reviews, 175 commits, 127 PRs in 2 years 8 months
Contributions summary:Vedanuj primarily contributed to the development and testing of utility functions and modules within the Pythia framework. They added unit tests for various utility functionalities including preprocessing, timer, and general utility functions, ensuring code quality and functionality. Furthermore, the user identified and addressed issues in model loading, fixed bugs related to data parallel checkpoint attributes, and added tests for the initialization of the processor in the base dataset class.
pytorchmodular-frameworkdialogvisionvision-language
facebookresearch/vilbert-multi-task
Aug 2019 - Jun 2020
Multi Task Vision and Language
Role in this project:ML Engineer
ML EngineerContributions:71 commits, 22 PRs, 22 pushes in 10 months
Contributions summary:Vedanuj implemented a script for extracting visual features from PyTorch models, likely for use in the vision-language tasks. They removed a Pythia dependency, indicating a refactor to reduce dependencies. The user added a variety of datasets, including those for Visual Genome QA, Visual Entailment, NLVR2, GQA, Guess What QA, and Flickr Grounding, demonstrating an ability to integrate different visual-language datasets. They also addressed issues related to data loading, distributed training, and evaluation of models.
multi-taskvisiondeep-learningcomputer-visionmachine-learning
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.