Ankita De
Software Engineer at Facebook AI
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolAnkita De is a software engineer with 11 years of experience based in the San Francisco Bay Area, currently contributing to Facebook AI on multimodal and conversational ML tooling. She has a strong academic foundation from BITS Pilani and UT Austin and a track record of shipping practical improvements to high-profile open-source projects like ParlAI and TorchMultimodal, including Hugging Face agent integrations, TorchScript support, and FLAVA checkpointing and metrics. Her work bridges research and production: she builds robust model-loading, debugging, and testing utilities that make large-scale pretraining and fine-tuning more reliable. Known for attention to reproducibility and engineering polish, she often focuses on subtle infrastructure details—path management, dictionary loading, and accuracy tracking—that materially improve developer experience and model reliability.
11 years of coding experience
1 year of employment as a software developer
BITS Pilani, Birla Institute of Technology and ScienceMaster of Science (M.S.), Computer Science, Master of Science (M.S.), Computer Science at The University of Texas at Austin
Github Skills (18)
transformers10
pytorch10
python10
parameter-tuning10
machine-learning10
multimodal10
huggingface-transformers10
trainings10
natural-language-processing10
fine-tuning10
bert10
transformer10
gpt10
nlp10
modeling10
Programming languages (2)
Github contributions (5)
facebookresearch/multimodal
Mar 2022 - Jan 2023
TorchMultimodal is a PyTorch library for training state-of-the-art multimodal multi-task models at scale.
Role in this project:
ML Engineer
ML EngineerContributions:267 reviews, 133 commits, 114 PRs in 10 months
Contributions summary:Ankita primarily contributed to the development of multimodal models using PyTorch. Their work involved creating and refining utilities for fine-tuning and pretraining FLAVA models. They introduced checkpointing, accuracy metrics, and resolved issues to ensure the models function correctly in both pretraining and fine-tuning scenarios. Additionally, the user implemented testing and debugging for the FLAVA model to validate and ensure its functionality.
pytorchmulti-taskartdeep-learningmultimodal
facebookresearch/ParlAI
Aug 2020 - Jul 2021
A framework for training and evaluating AI models on a variety of openly available dialogue datasets.
Role in this project:ML Engineer
ML EngineerContributions:2 reviews, 5 commits, 12 PRs in 10 months
Contributions summary:Ankita primarily contributed to the Hugging Face GPT2 and Bert agents within the ParlAI framework, enhancing their functionality and improving integration with the Hugging Face ecosystem. Their work involved modifying the agents' code to support loading pre-trained models from different directories and incorporating path management. These changes aimed at improving the flexibility of model loading and integrating with external resources for models. Furthermore, the user also modified the dictionary loading process and implemented TorchScript support for the greedy search module.
nlpdeep-learningdatasetmachine-learningtraining
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.