Roopali Vij
San Francisco Bay Area United States
Summary
👤
Senior🎓
Top SchoolRoopali Vij is an independent AI research consultant and seasoned LLM researcher with nine years of experience designing, evaluating, and deploying large-scale language models. At Google DeepMind she was a core contributor to Gemini and PaLM 2, leading pretraining data curation, long-context and multi-turn evaluation frameworks, and safety-aligned RL pipelines for production systems like Bard. She combines deep research rigor with production engineering—having captained pretraining runs, built large-scale data pipelines, and contributed RL data integrations to the widely used TF-Agents library. Based in the San Francisco Bay Area, she reviews for top conferences (NeurIPS, ICML, ICLR, AAAI) and is authoring two forthcoming books on generative AI, reflecting a strong commitment to community impact and knowledge sharing.
9 years of coding experience
5 years of employment as a software developer
Master of Science (MS), Computer Science, Master of Science (MS), Computer Science at University at BuffaloBachelor of Technology (B.Tech.), Electronics and Communications Engineering, Bachelor of Technology (B.Tech.), Electronics and Communications Engineering at Guru Gobind Singh Indraprastha University
English
Stackoverflow
449reputation
442kreached
23answers
0questions
Github Skills (18)
ai-agent10
python10
machine-learning10
reinforcement-learning10
agent10
tensorflow10
reverse10
algorithm9
algorithms9
data-pipelines9
data-preprocessing9
data-pipeline9
sql-server6
transactionscope6
entity-framework6
Programming languages (1)
Github contributions (5)
tensorflow/agents
Feb 2022 - Jul 2022
TF-Agents: A reliable, scalable and easy to use TensorFlow library for Contextual Bandits and Reinforcement Learning.
Role in this project:
ML Engineer
ML EngineerContributions:16 commits in 5 months
Contributions summary:Roopali primarily contributed to the TF-Agents library, focusing on the integration and enhancement of RLDS (Reinforcement Learning Datasets) data loading and processing capabilities. Their work included adding functionality to convert RLDS datasets into TF-Agents trajectories, integrating RLDS datasets with Reverb replay buffers, and modifying existing components like the CQL SAC and PPO Agent to use RLDS datasets. These changes indicate a focus on improving data ingestion and compatibility within the TF-Agents framework to enable the training of RL models using real-world and benchmark datasets.
scalabletf-agentsmultiagent-reinforcement-learningtensorflow-librarymulti-armed-bandits
roopaliv/os161
Jul 2017 - Jul 2017
Contributions:37 pushes, 1 branch in 1 day
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.