- Joined
- Jan 2025
- How they joined
- Directly
- Before Cognition
- Jump Trading GroupQuantitative Research Intern
- Status
- Current
“at cognition @moritz_stephan was the goat at delivering value. it was a blast working with him and I'm so excited to see what's next with Hone!”
- Credited as a co-author of SWE-1.7 ('Frontier Intelligence at a Fraction of the Cost') alongside Ray Wang, Gary Chang, Steven Cao and Silas Alberti (equal-contribution byline); SWE-1.7 targets longer-horizon asynchronous tasks.Jul 8, 2026
- Listed first in the author byline of Cognition's 'Introducing SWE-1.5: Our Fast Agent Model' (with Nikhil Jha and others), a frontier-size coding model served via Cerebras at up to 950 tok/s and launched in Windsurf; trained with end-to-end RL on the Cascade agent harness and beta-tested as 'Falcon Alpha'.Oct 29, 2025
- Gave the talk 'Cognition's DeepWiki' at the LangChain SF meetup on LLM wikis and agent memory (co-billed with LangChain's Brace Sproul), covering DeepWiki's move from an orchestrated v1 to an agentic v2, wiki-quality metrics and four context-engineering principles; DeepWiki grew from Devin context to 1.4M indexed repos.Jul 28, 2026
- On SWE-1.5 launch he posted: "The best way to make coding agents interactive is to just make them run faster than you can think. We're already hitting internal PMF with SWE-1.5, and this is only the beginning!" (via Techmeme, which carried the post).Oct 29, 2025
- Works on Research & Engineering at Cognition since January 2025 per LinkedIn ('Building Devin'); he joined about six months before Cognition agreed to acquire Windsurf (July 2025).Jan 2025
- MIT ICPC team 'Taxi' with Siyong Huang and Thomas Guo placed 18th at the ICPC 2023 World Finals (7/11 solved, held April 2024); the same trio competed at NAC 2022 as 'mit' and at the 2023 EC-Final as 'mit taxi'.Apr 18, 2024
- IOI record for Singapore (as Jacob Teo Por Loong / jacobtpl): Silver at IOI 2015 (rank 37), Gold at IOI 2016 (rank 11), Silver at IOI 2018 (rank 42).Aug 2016
- Promoted the talk on X: "excited to share about how we built DeepWiki at @cognition, and chat with @BraceSproul about the future of agent context and memory!"Jul 23, 2026excited to share about how we built DeepWiki at @cognition, and chat with @BraceSproul about the future of agent context and memory! rsvp here: https://t.co/ouxA2MWeEf
- On the SWE-1.7 launch day he posted on X: "Our best model so far and much more to come :) Excited for everyone to try it in Devin!"Jul 8, 2026Our best model so far and much more to come :) Excited for everyone to try it in Devin!
- Before Cognition, interned at Anysphere (Cursor) in January 2024, fine-tuning new models for Cursor Tab - early hands-on work on code-completion models two years before SWE-1.5.Jan 2024
- Jump Trading: Quantitative Trading Intern (Chicago, summer 2023) working on US equity microstructure, execution algorithms and HFT models, and winner of the Jump Global Quant Trading Competition Challenge; returned summer 2024 as Quantitative Research Intern doing alpha research on HF equities and mid-frequency options and building an options trade simulator.Jun 2023
- Research Science Institute 2017 researcher under Prof. Jelani Nelson (Harvard) on succinct data structures, producing arXiv:1709.09574 'Fillable arrays with constant time operations and a single bit of redundancy' with Jelani Nelson and Huacheng Yu.Sep 27, 2017
Building Devin
- Alpha research on high-frequency equities and mid-frequency options - Built an options trade simulator and a performant regression pipeline
- Building Cursor - Fine-tuning new models for Cursor Tab
- Training on simulated trading, predictive modeling and execution strategies - Projects on US equity market microstructure, execution algorithms and models for high-frequency trading. - Winner of the Jump Global Quant Trading Competition Challenge.
- Developed a new and improved algorithm for human location tracking through radio signals - Researched on removal of signal interference from noisy raw data
- Analyzed price mechanisms of and developed infrastructure to trade on automated market makers on the blockchain - Researched strategies for mid-frequency cryptocurrency quantitative trading
Added features to a talent pool portal used for company HR operations
TA for 6.172 Performance Engineering of Software Systems
Classes Taken: • CS5330 Randomised Algorithms • CS3244 Machine Learning • CS2100 Computer Organisation • CS2102 Database Systems • CS2106 Operating Systems • MA1101R Linear Algebra
insane how much our model training has progressed in less than a year. and we're only just getting started!
at cognition @moritz_stephan was the goat at delivering value. it was a blast working with him and I'm so excited to see what's next with Hone!
excited to share about how we built DeepWiki at @cognition, and chat with @BraceSproul about the future of agent context and memory! rsvp here: https://t.co/ouxA2MWeEf
@wowitsmrinal @cognition welcome, super excited to work with you!!
@jaredpalmer @cognition @asha_shar @ScottWu46 welcome!!
Our best model so far and much more to come :) Excited for everyone to try it in Devin!
seeing all the old house photos really brings me back. lots of fun memories and late night gym sessions... wish I took more pictures but all I have is us trying to recreate the cognition logo with dumbbells https://t.co/GF17c2X7Y1
@katiiecheng @imjaredz @silasalberti welcome and so excited to work together!!
@hone congrats!!!
the critical issue of our time fr https://t.co/VKJCCIOfYL
poetic that our catfishing trivia goat is now himself immortalized on wikipedia https://t.co/kZxv13e3Kn
@hone congrats!!
