Akshit Sinha is a machine learning researcher-engineer focused on evaluating and improving the capabilities of large language models, with a long-term aim of building AI that can autonomously perform complex, long-horizon tasks. Currently a Member of Technical Staff at Proximal, he develops post-training coding agents and reinforcement learning environments for extended coding workflows. He holds an MPhil in Machine Learning from Cambridge and completed research and internships at institutions including IIIT Hyderabad, University of Virginia, and Adobe, blending rigorous research with hands-on engineering. His undergrad work explored interpretability and machine unlearning for graph neural networks, and he has practical mobile development experience as a contributor to the widely used AnkiDroid project. Akshit’s profile reflects a rare mix of RL/coding-agent engineering, multimodality evaluation, and user-facing product improvements—suggesting an ability to bridge deep research with real-world tooling. Based in New Delhi with four years of experience, he often focuses on robustness and longevity of AI behavior rather than short-term benchmarks.
5 years of coding experience
3 years of employment as a software developer
Master of Philosophy - MPhil, Machine Learning and Machine Intelligence, Master of Philosophy - MPhil, Machine Learning and Machine Intelligence at University of Cambridge
Bachelor of Technology - BTech (Honors), Computer Science, Bachelor of Technology - BTech (Honors), Computer Science at International Institute of Information Technology Hyderabad (IIITH)
AnkiDroid: Anki flashcards on Android. Your secret trick to achieve superhuman information retention.
Role in this project:
Mobile Developer (Android)
Contributions:79 reviews, 34 commits, 28 PRs in 7 months
Contributions summary:Akshit contributed to the AnkiDroid Android application, primarily focusing on UI and feature enhancements related to the user interface and user experience. They addressed Kotlin migration issues by fixing typos and adding lint checks for correct tag formatting. Furthermore, they added a button for video clips in the note editor and improved the display of search results, including adjustments to the display of the average ease, interval, and other card/note stats in notes mode. These contributions demonstrate a focus on improving the application's user-facing aspects and code quality.
AnkiDroid: Anki flashcards on Android. Your secret trick to achieve superhuman information retention.
Contributions:12 PRs, 159 pushes, 47 branches in 11 months
androidanki
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.