JJJacob Jackson
Competitive programmer (IOI gold medalist for Canada 2014, 2015) turned ML tooling founder and researcher; creator of TabNine and Supermaven, ex-OpenAI Research Scientist, now at Cursor.
- Identity resolved: the Canada IOI 2014/2015 contestant 'Jacob Jackson' (handle zxqfl, IOI stats id 5053) is the same person as the Waterloo CS student who created TabNine and worked at OpenAI — his personal site's About page explicitly links both his competitive-programming profiles (codeforces zxqfl and stats.ioinformatics.org/people/5053) and his TabNine/OpenAI history. This resolves the caller's unverified lead handle 'taiga-206658396' and LinkedIn jacobbfjackson to a single individual.Dec 8, 2018
- Quant employment verification: CONFIRMED. Jacob Jackson (Canada IOI 2014/2015; University of Waterloo CS 2015-2019; LinkedIn jacobbfjackson) held two quantitative-finance internships in 2018 while an undergraduate: Software Development Intern at Jane Street (New York, 2018), where he designed developer-facing interactive UI software, and Winter Intern at Hudson River Trading (New York, 2018), where he used statistical models to extract predictive signals from market data and built a low-level C++ order-book representation. Jane Street is a proprietary trading firm/market maker and HRT is a quantitative trading firm, so both are genuine quant-finance employers (not merely names containing 'quant'); these are actual internships, not competitions, camps, or offers. Source: LinkedIn-derived enrichment (index-only for this timeline) corroborated by the firm websites' self-description; the identity link rests on the Waterloo-CS + IOI + TabNine/OpenAI corroboration on his personal site.
- Founded Supermaven (2023-2024), a fast code-completion product, which was acquired by Cursor/Anysphere in October 2024; he then joined Cursor as a researcher (2024-).Oct 2024
- Created TabNine, a deep-learning code-completion tool, working on it part-time while a Waterloo undergraduate (2018-2019); it was acquired by Codota in November 2019.
- Won IOI gold medals for Canada in both appearances: rank 15 at IOI 2014 and rank 9 at IOI 2015 (per the Competitive Programming Hall of Fame timeline).
- Interned at OpenAI in 2018 doing research in semi-supervised learning (arXiv 1902.02336), then returned as a Research Scientist at OpenAI from 2019 to 2022 before leaving to found Supermaven.
- Competed at the ICPC World Finals for University of Waterloo in 2016 (team placed 13th) and 2017 (18th), per CPHOF and Waterloo CS news.
- Earlier software-engineering internships while at Waterloo: MemSQL (San Francisco Bay Area, 2016), expanding the SQL compiler to run user code in-database, and Petuum Inc. (Greater Pittsburgh, 2017), designing an approximate clustering algorithm in a distributed-memory engine; the Petuum work became a SysML 2018 submission.
- Writes technical blog posts on his personal site, including 'Polyhedral Order Books' (2023), 'Gradients of GAN Objectives' (2021), and 'Searching for Style' (2020) — the order-book post indicates a continuing interest in market-microstructure mathematics.May 25, 2023
- Built Same Energy (same.energy), a visual search engine, as a personal project.
- Public X/Twitter handle @jbfja, linked from his personal site's About page; also reachable by email [contact omitted].
- Education: high school at University of Toronto Schools (2011-2015), then a Bachelor's degree in Computer Science at the University of Waterloo (2015-2019).
Supermaven was a code completion product acquired by Cursor in October 2024: https://www.cursor.com/blog/supermaven
San Francisco Bay Area
I created TabNine, the first commercial tool to use GPT-2 to write code, and worked on it part-time while in college. TabNine was acquired by Codota in November 2019. https://tabn
Research in semi-supervised learning. https://arxiv.org/abs/1902.02336
Designed software providing an easy way for other developers to create interactive user interfaces
Used simple statistical models to extract predictive signals from market data Applied low-level optimizations to create a performant order book representation in C++
A submission based on this work was accepted to SysML 2018: http://www.sysml.cc/doc/105.pdf Designed and implemented an approximate clustering algorithm in a distributed memory en
Competition record
Canada · IOI