The Minds Behind the Models: Taro Kogawa
Taro Kogawa: The High School Graduate Who Let AI Do the Coding
Taro Kogawa from Tokyo just graduated from high school. He has never formally learned to code, has no background in data science, and hasn't studied time‑series modelling. Yet he managed to place in the top 200 of the IMC Prosperity 4 challenge (out of 22,000 participants), reach 4th place in Cambridge Battlecode 2026 (out of 500), and make submissions to AlphaNova that survived the full gauntlet of overfitting and correlation filters, with two of his signals in the top 20 - a rare feat only two participants in the top 20 managed to do!
His secret is not a supercomputer or an expensive API key. It's a $20‑per‑month Claude Pro subscription and a relentless focus on abstract problem‑solving.
“To me, using AI doesn’t mean simply using a powerful model or paying for an expensive plan, consuming huge amounts of tokens, and leaving everything to the AI. The important part is to take the abstract thinking and problem‑solving skills that I’m good at and communicate them to AI through prompts, while letting the AI handle the concrete implementation and tasks that require specialized knowledge.”
Taro’s path into quantitative finance was straightforward: he was drawn to the intersection of mathematics, markets, and competition. After becoming a finalist in Cambridge Battlecode - a fast-paced, resource-constrained coding challenge that heavily overlaps with the problem-solving style valued in quantitative finance - he gained direct exposure to the industry and decided to pursue it further. Along the way, he also became a finalist in the Japan Mathematical Olympiad 2025, placing around 30th out of 5,000 participants.
“I’ve always been interested in three things: mathematics, markets, and competition. Quantitative finance was a field that brought all three of these together, so I naturally became interested in it.”
His AlphaNova workflow is a tight feedback loop between AI‑driven analysis and human judgment. He doesn't write much code himself. Instead, he describes the hypotheses he wants to test, has the AI generate the implementation, examines the resulting signals and their behaviour, and then gives further instructions.
“I have the AI analyze the data, examine the hypotheses and signals it comes up with, and then give that feedback back to the AI. I repeat this process to search for better signals.”
This approach has allowed him to participate at a high level without the traditional years of programming and data‑science training. He's also applying the same method to other projects: developing high‑frequency trading strategies (including market making in prediction markets) and longer‑term crypto investment models, while competing in the Florent Code League.
When asked what tools he reaches for, Taro’s answer is characteristically honest:
“To be honest, I rely quite heavily on automated analysis by AI, so I don’t actually know exactly what libraries I’m using all the time. At least I’m pretty sure I’m using Python, haha.”
He’s also excited about the future of AlphaNova. When asked about the new biweekly competitions, he saw them as a levelling force:
“I’m very excited about AlphaNova’s new biweekly competitions as well. I think having many short competitions creates more equal opportunities for participants and makes the competitive aspect even more enjoyable.”
Taro Kogawa is proof that the barrier to entry in quantitative finance is shifting. When the bottleneck is no longer implementation speed but the quality of your ideas and your ability to direct AI tools, a high school graduate with strong mathematical intuition can compete alongside PhDs and industry veterans.
Stay tuned for more profiles from the AlphaNova community. If you’d like to be featured, reach out—we’d love to share your story.
Join the latest AlphaNova competition and see if your signals can survive the same rigorous tests that Taro’s did.