Founder

Sergiy Sanin

Sergiy Sanin

Twenty years building software, sixteen of them in algorithmic trading. Nine years of exchange connectivity, order routing and low-latency execution infrastructure at Credit Suisse, Standard Chartered and Tower Research; four years owning the core execution system at a Singapore crypto prop and market-making firm.

Since 2025, market-making on own capital under Delta Tech Consulting FZ-LLC — research, execution engines, risk and live operations, end to end. The measurement discipline that comes with running a book you own is the reason the research on this site reports the results that went against it as readily as the ones that did not.

Track

2025 —

Delta Tech Consulting FZ-LLC

Founder · Quant Trader-Engineer

Market-making on own capital. Two independent live systems: a paired spot+perp delta-neutral maker and a single-leg perpetual maker, with the complete stack behind them — Rust execution engines, market-data recorder and deterministic replay, a queue-aware fill simulator validated against private fill tapes, parameter fitting with out-of-sample gating, per-leg P&L attribution, and automated kill-gates.

2021 — 2025

Tokka Labs · Vega Solutions · Altonomy

Trading Strategy / Quant Developer · Singapore

Owned the core execution system — the primary hedging engine and the arbitrage execution layer — at eight-figure notional. RFQ quote generation, firm execution and post-trade hedging across 1inch, 0x and Hashflow. Designed and implemented the price-aggregation and oracle system for Pyth Network.

2019 — 2021

Tower Research Capital

Software Platform Architect · Singapore

Medium-frequency trading platform inside a pod trading commodity futures. Order-flow handling, risk checks, recovery.

2011 — 2019

Credit Suisse · Standard Chartered

AVP, Electronic Trading Systems · Hong Kong / Singapore

Exchange connectivity, order routing and execution-flow handling for market-making and execution workflows.

Published

Working paper — RL over a calibrated market-making baseline learns a constant, not a policy. A pre-registered experiment on my own market maker in which every discriminating prediction failed, and folding each policy’s mean action into a single constant then recovered 132% of the trained policy’s edge and beat it in fourteen runs of fifteen. Roughly 71% of the measured improvement is accounted for by the policy simply trading less. The paper also withdraws a conclusion from earlier work of mine that failed to replicate, and lists ten defects found in my own code.

SSRN · data and verifier — over 300 assertions, non-zero exit on any disagreement.

Shorter measurement notes — cross-venue latency, markout against rebate, simulator fidelity, sample-size discipline and the economics of retail market making — are collected under research.

Engagement

Delta Tech Consulting FZ-LLC holds a RAKEZ Services licence (Software House; Computer Systems and Software Designing) and contracts on a B2B basis — trading-systems development, execution analytics, venue integration and strategy R&D for institutional counterparties. Remote, or Singapore and Hong Kong.

Contact

Telegram
@ssanin82