The Minds Behind the Models: Vivaan Jain

Vivaan Jain: The Olympiad Mindset and the Pre‑Submission Gauntlet

Vivaan Jain is a Computer Engineering student at Dwarkadas J. Sanghvi College of Engineering in Mumbai, pursuing an honors degree in Data Science and serving as an ML Mentor. His background is steeped in AI—LLMs, agentic systems, and RAG—but his entry into quantitative finance started much earlier, with pure mathematics.

"It started with mathematics. I spent my school years on Olympiads and won a bronze medal at the International Mathematics Olympiad. Math taught me to love problems that do not give up their answers easily."

That attraction to difficult problems naturally found its way to the markets. But it wasn't the allure of money that hooked him; it was the puzzle.

"In high school I was introduced to the stock market and started trading a little. What hooked me was not the profit. A market looked like the hardest unsolved problem I had ever seen, with a live scoreboard attached."

To bridge the gap between Olympiad math and live markets, he started automating strategies and watching them fail. Each failure became a lesson in statistics and market structure. It is a familiar origin story for the best quants: the market is a ruthless teacher, and the only way to survive is to learn how to listen to it.

Building the Gauntlet: "Assume Your Best Result is Partly Luck"

When asked about his most valuable tool, Vivaan didn't name a library or a model. He named a piece of software he built entirely from scratch.

"My most valuable tool is one I wrote myself: a pre‑submission tester that every strategy has to pass before it leaves my machine. I built it from scratch in plain numpy and pandas, so I know exactly what every check does."

This custom validator is designed to be brutally honest. It cuts off the future and scrambles it to prove there is no leakage. It retrains on shuffled data to see if an edge is real or just overfitting. It uses block bootstrapping to measure long-history performance so a single lucky streak cannot flatter a signal. It even measures how similar a new idea is to his older ones.

"The principle behind all of it: assume your best result is partly luck, and build the tools to find out how much."

This rigorous skepticism paid off in his external research on Indian options markets, where he built a pricing model to account for shifts in risk when exchanges changed option expiry days.

"When I tried to prove the change caused the shift, my own pre-agreed test said I could not, and I accepted that... The biggest lesson was about myself: be most suspicious of the results that agree with what you hoped to see."

Winning The Third Biweekly Competition: Trading the Speed, Not the Level

Vivaan's approach to the AlphaNova competition was highly iterative, and his path to winning our third biweekly competition was built on a series of instructive failures.

He started with the textbook approach—short-term reversal sized by shrunk covariance—but realized it was too close to what everyone else was doing. The competition rewards novelty. So he pivoted to the other extreme, building a complex blend of path signatures and statistical arbitrage blocks.

"It lost money. That was the lesson I needed, because the recent window I was using to judge it had been telling me the opposite."

He realized his measurement framework was flawed. He went back to the drawing board, moved to long-history testing, and wrote his own checks. He also made a crucial mistake that many quants make when trying to pass a correlation filter: he added artificial "filler" to make his signal look different. His own overfitting gate rejected it immediately.

"I wrote down 'never again' and I meant it."

His breakthrough—the insight that won Cycle 3—came from changing what he traded.

"Instead of the level of a signal, I traded how the same behaviour looks at a fast speed against a slow one. That cancels the persistent bias that makes so many entries look alike."

By trading the spread between fast and slow behaviors, he stripped out persistent market bias. He then layered on careful execution logic—chasing signals that were strengthening, letting decaying ones fade slowly, and slowly removing per-asset drift.

"Every step was tested against the one before it, and I treated each small gain as noise until it proved otherwise."

AI as Execution, Not Ideation

Vivaan uses AI heavily, but his division of labor is absolute.

"Heavily, but the thinking stays with me. I do the research. The idea, the hypothesis and the plan for testing it are mine. I do not ask artificial intelligence to find alpha."

He uses the AI as an execution engine inside strict rules he sets. The assistant builds the strategy, but nothing is accepted until it passes his custom pre-submission tests. He uses AI as a sparring partner to ask "why am I seeing this, and what could be better?"

"Repetitive implementation is better done by machines, so humans can spend their time on ideas and research, where the real edge is made."

It is a perfect distillation of the shift we've seen across this series. The tools have made implementation cheap, but the judgment—knowing what to build and what to throw away—remains strictly human.

What We Can All Learn

Vivaan’s journey to winning Cycle 3 offers three powerful lessons:

Build your own gauntlet. Don't rely solely on the platform's tests. Build a pre-submission tester that scrambles data, blocks bootstraps, and checks for leakage so you know exactly what your signal is doing.
Trade the spread, not the level. Persistent biases ruin naive signals. Looking at how a behavior changes over different speeds (fast vs. slow) can cancel that bias and reveal true predictive power.
Be most suspicious of your best results. The results that perfectly align with your hopes are the ones most likely to be overfit. Treat every small gain as noise until it proves otherwise.

That combination of Olympiad-level rigor, engineering discipline, and deep intellectual honesty is exactly what the AlphaNova community is built on.

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.