He Jia (贾赫)

Managing Director and Quant Research Lead at Capibara Capital

Introduction — Hello! My name is He Jia (贾赫, pronounced as /hɜ: dʒɑ:/ in Chinese, which roughly means famous businessman 😅). Born in Jingmen in 1998, I moved to Shenzhen with my family in 2008. I then went to Beijing for college in 2016 and Princeton for graduate school in 2020 (physically in 2021 though, due to the pandemic).

I was on the 2024/25 astrophysics postdoc job market, but I made a last-minute decision to turn down my postdoc opportunities and leave academia, for now, to join Capibara Capital, a quantitative trading startup. I defended my PhD in Astrophysics on July 29, 2025, with Prof. Eliot Quataert as my main advisor.

I am currently a Managing Director (MD, I have only one doctorate degree) and Quant Research Lead at Capibara Capital, where I oversee deep-learning model training across hundreds of GPUs to generate predictive trading signals. Our multi-billion-dollar book spans both MFT and HFT strategies across mainland China, Hong Kong, Japan, and South Korea, and covers equities, futures, and options. I also co-lead the portfolio optimization team.

Academic Career — I am still a somewhat active, part-time astrophysicist, broadly interested in using advanced statistical techniques to optimally extract quantitative information from astrophysical data. While I was initially trained as a cosmologist, a field I continue to contribute to, my research has expanded to include black holes since graduate school. Beyond cosmology and black holes, I am also open to exploring new areas wherever the most exciting scientific opportunities arise.

I was a visiting undergraduate researcher at UC Berkeley during parts of 2018 and 2019, working with Prof. Uroš Seljak. Unfortunately, I still owe him a BayesFast paper. This paper never came out because I wanted to release it together with the code. Back in 2020, however, there was no suitable autodiff framework available for my purposes, so I ended up hand-writing my own, which was later criticized for not being very user-friendly. By the time JAX significantly improved its latency in 2022, Uroš’s group had already moved on to developing more general methods, although I believe BayesFast remains the most efficient option for simpler problems.

Beyond Work — I am a speed enthusiast with a passion for aviation and auto racing. As a dual-rated FAA private pilot with the high-performance endorsement, I fly airplanes as well as helicopters. Aircraft I have flown include Cessna 172, Diamond DA40 (my favorite airplane so far), Cirrus SR20, Robinson R22, Robinson R44, and Cabri G2 (my favorite helicopter so far). When I want some fun on the ground, I enjoy driving cars like Tesla Model 3 Performance, Xiaomi SU7 Ultra, Porsche 911, and various Ferrari models on race tracks.

I play tennis and am currently self-rated NTRP 3.5. My favorite band is The Landlord’s Cats. I am a strong believer in techno-optimism. I plan to establish the Jia Foundation in the not-too-distant future to support the scientific fields I have long been passionate about.