Khurram Javed
Khurram Javed is a reinforcement-learning researcher and a co-founder of Oak Lab. After competing in mathematics and studying software engineering in Pakistan, he worked with Martha White and then Richard S. SuttonRichard S. SuttonRichard S. Sutton is a Canadian computer scientist and professor at the University of Alberta. He developed temporal-difference learning to explain how agents revise their predictions from the consequences of their actions. He and Andrew G. Barto wrote Reinforcement Learning: An Introduction and shared the 2024 A.M. Turing Award.Open the full entry at the University of Alberta in Canada. His research concerns how systems with limited computing resources can learn from a continuing stream of experience.
Contents6 sections
Key facts
2012–2018Mathematics competitions and software engineering
From 2012 to 2014 Javed attended International Mathematical Olympiad training camps at the Abdus Salam School of Mathematical Sciences in Lahore, Pakistan. He received an honorable mention at the 55th International Mathematical Olympiad and a bronze medal at the 26th Asian Pacific Mathematics Olympiad. In August 2014 he enrolled in software engineering at the National University of Sciences and Technology in Islamabad, studying under Faisal Shafait. He graduated in July 2018 with a 4.0 GPA, the President’s Gold Medal for the highest-ranked graduate and the Rector’s Gold Medal for best thesis.[1]
In 2017 he contributed to introductory machine-learning teaching at his university and spent June through September at EPFL. There Patrick Thiran supervised his research on assessing machine-learning methods for network routing. His early papers with Shafait addressed locating documents in natural images and learning new classification tasks incrementally.[1]
2018–2020Graduate studies in Canada
He began an MSc at the University of Alberta in Edmonton in September 2018, supervised by Martha White. The two published on learning representations for continual learning at NeurIPS in 2019. From June to October that year he worked at Huawei Research’s HiSilicon on the use of reinforcement learning in chip design. He was a visiting researcher at Mila in Montreal from February to June 2020, working with Yoshua BengioYoshua BengioYoshua Bengio is a professor of computer science at Université de Montréal, founder of Mila and a deep learning pioneer who shared the 2018 Turing Award with Geoffrey Hinton and Yann LeCun; since 2023 he has turned to AI safety research, chairing the International AI Safety Report and founding the non-profit LawZero in 2025. He appeared on the Lex Fridman Podcast in 2018.Open the full entry on discovering causal relationships from observations arriving over time. He completed his MSc that September with the thesis Learning Online-Aware Representations using Neural Networks.[1][2]
2022–2025Turning a collaboration into a doctorate
After his MSc, Javed intended to enter industry. A discussion drew him into a project with Richard S. Sutton. About six months of collaboration persuaded them to turn it into a doctoral topic. He began his doctorate under Sutton in January 2022 and completed it in February 2025. His dissertation, Real-time Reinforcement Learning for Achieving Goals in Big Worlds, focused on learners updating their decisions and pursuing goals in real time when their computational capacity is limited. He also contributed to teaching reinforcement learning at Alberta.[1][3][4]
In 2024 his SwiftTD paper with Arsalan Sharifnassab and Sutton received a best-paper award at the Reinforcement Learning Conference. That October he joined the small research team at Keen Technologies, working with John CarmackJohn CarmackJohn Carmack is an American programmer and computer-game designer, a co-founder of id Software who led the technology behind first-person shooters such as Wolfenstein 3-D, DOOM and Quake; he later served as chief technology officer of Oculus and in 2022 founded the artificial general intelligence company Keen Technologies.Open the full entry and Sutton. Their 2024 workshop paper set out the big world hypothesis, that an agent with finite computational resources faces an environment vastly more complex than itself. In 2025 Javed presented his own Swift-Sarsa work on fast, robust linear control.[1][5]
August 18, 2026Oak Lab and continual learning
Javed and Sutton co-founded Oak Lab to study systems that change as they gain experience. In Sequoia’s interview, around 23:40, Javed pointed out that even extensively trained models need not learn from the events a user encounters after deployment if their parameters remain fixed. New research directions, he added, often perform worse than mature approaches at first. A team that insists on immediately beating existing benchmarks risks abandoning a promising direction too early.[4][6]
Related Organizations, Websites and Public Accounts
EPFL: Former research intern. Official website: https://www.epfl.ch/ (opens in a new window).[1]
HiSilicon: Former research associate. Official website: https://www.hisilicon.com/ (opens in a new window).[1]
Mila: Former visiting researcher. Official website: https://mila.quebec/ (opens in a new window).[1][2]
Keen Technologies: Former research scientist. Official website: https://keenagi.com/ (opens in a new window).[1]
Oak Lab: Co-founder. Official website: https://oaklab.ai/ (opens in a new window).[4][6]
Personal website: https://khurramjaved.com/ (opens in a new window). He also helped organize the RL in Big Worlds research event.[5][7]
LinkedInLinkedInLinkedIn is a social networking platform for professionals. It was founded in 2002 by Reid Hoffman and others, launched in 2003, and listed on the New York Stock Exchange in 2011; after Microsoft acquired it for approximately $26.2 billion in 2016, it became a wholly owned Microsoft subsidiary. Its revenue comes from membership subscriptions, advertising, and recruitment solutions.Open the full entry: https://www.linkedin.com/in/kjaved2/ (opens in a new window).[5]
Sources7 sources
- Khurram Javed: Curriculum vitae (opens in a new window)khurramjaved.com↑
- Khurram Javed (opens in a new window)mila.quebec↑
- Real-time Reinforcement Learning for Achieving Goals in Big Worlds (opens in a new window)ualberta.scholaris.ca↑
- Rich Sutton and Khurram Javed: Why AI Models Stop Learning, and How to Start It Again (opens in a new window)youtube.com↑
- Khurram Javed (opens in a new window)khurramjaved.com↑
- The Big World Hypothesis (opens in a new window)oaklab.ai↑
- RL in Big Worlds (opens in a new window)rlinbigworlds.ca↑