Careers · Coimbatore, India
MarvelQuant is a proprietary trading firm. We research, test and run systematic strategies with our own capital, from the Indian open to the US close. We hire people who test their own ideas hard and want to see them trade.
Three teams share one trading system and work closely with each other. A strategy developer sees how their idea executes. An engineer sees the trades their code sends.
Write strategy logic for Indian index derivatives and other markets, backtest it against real costs and exchange rules, and compare live results with the tests. Some of the team build machine-learning models on order-book and options data.
Work on the trading system in Rust and the research and data tools in Python: market data records, broker connections, simulation and monitoring.
Supervise live strategies through their sessions, watch positions and margins, and reconcile trades with brokers.
We hire graduates and experienced people. Experience in finance helps. It is not required.
Turn a market idea into strategy logic with defined entry, exit and risk rules, then backtest it and track it live.
Build statistical and deep-learning models on order-book, trade and option-chain data, and judge them on periods they have never seen.
Build and run the systems that record market data, simulate strategies and trade them across Indian, US and crypto sessions.
Process and check market data, maintain databases and build the internal tools the strategy and operations teams use every day.
Supervise live strategies, monitor positions and margins through the session, and reconcile trades with brokers.
No role above fits? Tell us what you would work on and why it matters to a trading firm.
Send one email. Put the role in the subject line.
One or two pages. Tell us what you built or decided yourself.
Code, a notebook, a backtest write-up or a report you are proud of. Tell us what went wrong in it as well as what went right.
A few lines on the role you want and why.
Applications go to the firm’s administration desk, which passes them to the team that is hiring.
We never charge candidates a fee at any stage.
MarvelQuant never asks candidates for money, deposits or trading capital, and never recruits through trading groups or messaging channels. Treat any job offer that does as fraud.
No prize and no need to send it in. If you enjoy it, mention it when you apply.
A strategy wins 55% of its trades. Every trade wins or loses the same amount, and trades are independent of each other.
After 100 trades, what is the probability that the strategy is behind?
And after 400 trades?
The strategy is behind when it wins 49 trades or fewer out of 100. With a win rate of 0.55, that probability is about 13.5%. It finishes exactly level, 50 wins and 50 losses, another 4.8% of the time.
After 400 trades it is behind when it wins 199 or fewer. That probability falls to about 2.0%.
A normal approximation gets close. Over 100 trades the expected number of wins is 55, with a standard deviation of about 5, so finishing level at 50 wins sits 1 standard deviation below the mean. Over 400 trades the expected lead is four times larger and the standard deviation only twice as large, so level sits 2 standard deviations out.
A real edge loses money over 100 trades about one time in seven. This is why we judge strategies over long samples and never on a good month.
The strategy team works hybrid, from Tumakuru, Karnataka and from the Coimbatore office. The founder works from the Bengaluru office.