SigmaFi is a technology-driven quantitative trading firm. Founded by former Jane Street traders, Getco low-latency engineers and Google systems builders — we trade ETFs, equities, options, futures, fixed income and prediction markets from Hong Kong.
* Composite illustration of stylised research signals. Not an offer, recommendation, or claim of past performance.
§ 01 — Approach
Our research blends ideas refined at the world's top quantitative firms with a modern ML stack and a self-improving development engine we call CWS. Every strategy ships through the same risk, execution and monitoring substrate.
Cross-sectional pricing relationships, rank-based long/short books and rapid cross-market execution — capturing the mean-reverting noise between correlated instruments.
Relative-value signals that buy strength and fade exhaustion. Built across equities, ETFs, futures and digital assets, sized by realised volatility regimes.
A library of factor exposures — momentum, value, carry, low-vol, quality — combined under a risk-aware optimiser that rebalances daily across regions.
Custom matching-engine adapters, kernel-bypass networking and co-located gateways harvest fleeting price differentials across exchanges and dark pools.
Order-flow imbalance, options skew, news embeddings and on-chain signals fused via gradient-boosted ensembles and transformer-style sequence models.
+ N
New strategy hypotheses graduate from the research pipeline weekly — every promising signal is paper-traded, stress-tested, and only then promoted to production.
View research →§ 02 — Platform
Coding Workflow System turns research into a continuously self-improving loop. Large language models propose, test and refine trading strategies through a harness-engineered pipeline; the same strategy graph is replayed across markets and asset classes.
Harness-engineered LLM agents author signals, run backtests, and iterate — compressing weeks of quant grunt-work into hours.
Petabyte-scale market and alt-data pipelines, GPU/CPU cluster scheduling, deterministic backtest reproducibility.
Papers in, alpha out. Reinforcement-learning loops, cross-sectional and time-series template strategies validated continuously across markets.
σ ▸ cws propose --hypothesis "intraday momentum × news-tone" → 14 candidate signals generated · seeded by 132 papers → 3 promoted to vector backtest σ ▸ cws backtest --grid us-eq,hk-eq,jp-eq --window 5y [OK] us-eq · Sharpe 2.4 t-stat 6.1 [OK] hk-eq · Sharpe 1.9 t-stat 4.8 [KO] jp-eq · capacity bound — escalate σ ▸ cws stress --shock 2008,2015,2020,2022 → drawdown profile within risk budget → tail-correlation w/ existing book: 0.18 σ ▸ cws promote --to paper-trade --size $2.0M → promoted · monitoring: dash/sg-1421
Hypotheses / wk
~120
Backtests / day
~4,000
Strategies live
growing
§ 03 — Markets
Multi-asset coverage from Hong Kong as our anchor. Every venue we touch is governed by the same risk substrate and the same execution stack.
Liquidity provision and statistical strategies across US, Hong Kong, Japan and pan-Asia listings. Inheriting the ETF-pricing discipline our team built at Jane Street.
Volatility arbitrage, calendar and skew trades, listed-vs-OTC dislocations — backed by ML-driven implied-surface calibration and risk-parity sizing.
Yield-curve trades, basis swaps, cross-currency basis and macro spread strategies — the playbook refined at Citadel applied to a modern compute stack.
We started as a crypto market maker. We still trade BTC, ETH and majors, and we're pushing into prediction markets where short-horizon information edges abound.
§ 04 — Risk & compliance
A Citadel-style central risk function: real-time exposure across strategies, books and assets — with hard limits, kill-switches and per-desk PnL isolation.
Type 9 (Asset Management) licence application in progress with the Hong Kong SFC. Processes built to the standards of leading global financial centres.
We treat market integrity as a prerequisite, not a tax. Strategy review, surveillance and venue-level controls are first-class engineering concerns.
Multi-region disaster recovery, deterministic replay of every order ever sent, and independent infrastructure for compliance and audit.
§ 05 — People
Our founding team trained at Jane Street, Citadel, Getco, SIG and Google, with research roots at Princeton, MIT, Berkeley, Columbia and Tsinghua. We pair that lineage with a willingness to throw out received wisdom whenever the data says so.
About SigmaFi →Affiliations of current and former team members. Listed firms are not investors in or partners of SigmaFi.
§ 06 — Get in touch
For trading partnerships, capital introductions, and engineering recruiting — reach the team directly.
