About Poly Syncer
Poly Syncer powers automated Polymarket trading by mirroring the trades of the most profitable Polymarket wallets for our subscribers, in seconds, without holding their funds. The product is only as good as the wallets it picks. Our scoring model — described in the whitepaper — decides which wallets earn a leader slot, how much capital our subscribers route through them, and how quickly we de-list a wallet that has gone cold. The math is the moat.
The role
You will own the leader-trust scoring model and the Kelly-cap calibration that sits on top of it. The current model is a calibrated logistic blend of risk-adjusted return, hit rate, edge consistency, and drawdown recovery, with a discount factor that reflects market regime. It works. We have receipts. It is also not the model we will be running in twelve months — there are at least three open research questions where the literature does not give us an off-the-shelf answer.
You will have access to four years of fully labeled Polymarket fills (roughly 18 million events), a backtesting harness already built out in Python, and the engineers who wrote it. You will also have access to the live execution data from the past nine months, including counter-factuals from the policy layer (trades we declined and what would have happened). The infrastructure is in place. What we need is research judgment — the kind that decides which questions are worth answering, which results survive an honest stress test, and which deserve a page in the next whitepaper revision.
Critically, this is not a pure-research role. Research that does not ship is not finished. You will write production-grade Python or Go, your work will be code-reviewed by the engineering team, and your model will run in front of subscribers within a quarter. We have no patience for "we'll productionize it later." We will, however, give you the runway to get the research right before we ship.
You'll be a fit if
- You have a background in statistical modeling at a quant fund, market maker, crypto trading firm, or a comparable serious operator. Academia counts if the academic was honest about out-of-sample.
- You can write production-grade Python or Go — not just notebooks. Code review is real here.
- You have deep, instinctive skepticism toward your own backtests. You have been burned by overfitting and can tell the story of how.
- You understand the difference between calibration and discrimination, and you have an opinion on which matters more in our setting.
- You can read the whitepaper and tell us, on the intro call, what you would attack first.
- You write clearly. Research findings get circulated as memos here, not as PowerPoint.
- You are comfortable being wrong in writing. Half of research is publishing the negative results so the next person does not repeat them.
Bonus points
- You have published applied research on prediction markets, sports betting markets, or crypto market microstructure.
- You have worked on a wallet-clustering or address-attribution problem before.
- You have implemented Kelly-cap or fractional-Kelly logic in production at a real firm.
- You have an opinion on Bayesian model averaging that you can defend with data.
Process
- Intro call (45 min). A conversation with the head of research and the CEO. We talk through your background and our open questions.
- Paid take-home (12–16 hours, $1,800). A self-contained research problem on a sample of our data, with a written deliverable. You keep the work and the IP regardless of outcome.
- Research deep-dive (90 min). Two of our researchers walk through your take-home with you, then we walk through one of our internal memos.
- Founder call + offer (60 min). Compensation, equity, and the values described in the manifesto. Offer within five business days.
Compensation
$160k–$220k base + meaningful equity + performance component. The performance component is paid annually and tied to model live performance, not to model backtest performance — we do not reward optimistic backtests. Equity vests four years with a one-year cliff. Health, off-site, and co-working stipend as for the rest of the team.
Location
Remote-first. HQ is in Stockholm. We hire across EU and US time zones with at least four hours of overlap with European working hours, so research reviews can happen live when needed.
How to apply
Email [email protected] with links to research you have written — published papers, internal memos that have been shared externally, or a Kaggle write-up that demonstrates how you think. A short note on which of our open questions interests you most is welcome.
For technical context, the whitepaper and the methodology page are the canonical starting points. For company posture, see the about page and the manifesto.