APPLIED AI · CAPITAL MARKETS

Justock

An operations platform for a systematic trading desk — consolidated market intake, reproducible backtests, portfolio monitoring and a full audit trail.

YEAR

2026

INDUSTRY

Capital markets / systematic trading

SERVICE USED

Applied AI, deterministic pipelines, data engineering

CHALLENGE

Iterate strategy variants without losing the audit trail

(QTF® — THE PROBLEM)

Systematic desks live and die on the ability to iterate. But iterating fast is only useful if every run is reproducible, every position is monitored in real time, and every decision has a paper trail a compliance officer can follow. Most desks give up one of those three to move fast — which is expensive when it catches up with them.

Justock runs systematic strategies across Indian and cross-border equities. The operating problem was not signal quality — it was the tail: reconciling positions, replaying backtests, exporting reports for compliance, and answering the question ‘why did this trade fire on Tuesday’ three days later without a real answer.

One market-data pipeline

We consolidated the desk’s price, corporate-action and reference-data feeds into a single append-only store with cryptographic hashing at intake. Every downstream computation — backtest, monitor, report — reads from that store, never from a shared spreadsheet or a portable drive. Data provenance is queryable back to the tick.

Reproducible backtests

Strategy runs are deterministic: same code, same data snapshot, same random seed → same output, byte-for-byte. Every backtest is stamped with the git commit of the strategy, the hash of the data window, and the exact rule-pack version used. A backtest from six months ago replays identically today.

• Deterministic seeds and pinned data snapshots on every run.
• Strategy variant diffs preserved as first-class objects — every ‘what if we changed the stop by 20 bps’ is a queryable artefact.
• Portfolio-level risk gates evaluated on every proposed order — not after the fact.
• Compliance-ready audit trail with signed decisions and rationale text linked to the underlying signal.

Live monitoring, not spreadsheets

Positions, exposures, drawdowns and gate breaches surface on a real-time dashboard with the same math as the backtest engine. No end-of-day spreadsheet reconciliation — the numbers a trader sees at 14:37 are the numbers the risk officer sees at 14:37.

A strategy is only interesting if you can replay it. Everything else is a story.

What changed

The desk iterates strategy variants roughly 8× faster than before — measured over 90 days of comparable research volume — because every run is a one-line invocation and every result is a permanent artefact. Reconciliation windows dropped to zero because there is one source of truth. Compliance queries that used to take a week now take an afternoon.

(QTF® — CLIENT REVIEW)

The compliance auditor asked us why a trade fired six months ago. We showed her the signal, the rule-pack version and the exact data window in under five minutes. That was the moment I trusted the system.