ds↗
Recruiter brief
Quantitative financeSeptember 2026

Quantitative Finance & Markets

Prediction-market execution systems

Order reconciliation, execution controls, queue-aware research, and an independently packaged refactor of a prediction-market system.

PythonReconciliationEvent-driven systemsOffline testingQueue-aware replayQwenOpenAI-compatible inference
Built systemRank 01 of 12 in this field ↗
Why it sits here. The strongest execution-systems evidence: explicit order lifecycles, failure handling, and independent verification records.
06 / Two sides. One uncertain future.STUDY IN SPACE

THE QUESTION

What happens when an exchange accepts an order but the acknowledgement disappears?

Separate execution, experiments, and runtime state so a trading system can be inspected and recovered without treating a plausible backtest as proof.

THE MISSING RESPONSE
order intent
response lostunknown ≠ rejected
reconcile before retry
One intent.One accountable outcome.
Concept diagram · not performance data

01 / IMPLEMENTATION & CONTRIBUTION

What the work involves

Donald’s agent-assisted independent refactor separates execution, tape, backtesting, research, and dashboards, with explicit runtime paths and write controls. Earlier system and research snapshots retain shared-source attribution.

Technical depth

Partial fills, unknown order intents, restarts, and cancellation ownership all have to preserve quantity-level exposure.

Data and assumptions

Source snapshots, offline fixtures, request/response scenarios, and a separate research protocol. Private orders and account data are not exposed here.

A missing response is an unknown outcome. Passing simulated-client tests does not establish identical live behavior.

The project family

kalshibotkalshibot-membranekalshibot-membrane-crosssectionkalshibot-membrane-sweepfreeze

02 / THE EXPERIMENTAL RECORD

What the experiment taught

A clean refactor and a large test suite still cannot prove production equivalence. That acceptance gap remains visible in the project record.

What changed

The large live module was separated by responsibility. Ambiguous order creation is reconciled instead of automatically retried.

03 / RESULTS

What came out of it

An independently packaged execution/research system with reconciliation for ambiguous order outcomes. Dated September 4 verification records 1,526 passing offline tests; production equivalence and profitability are separate, unproven outcomes.

The lesson in the work

Separate implementation safety, execution fidelity, and economic validity into different acceptance gates.

04 / SUPPORTING EVIDENCE

Follow the source

Implementation notes, project records, and supporting artifacts.

Source context & project scope

Independent refactor in Donald’s project collection. Source and Git provenance checked; agent assistance and upstream boundaries retained.

An independent source snapshot separates execution, research, and mutable runtime state. Structural equivalence does not establish production equivalence.

REPORT · 2026-09-04

A non-idempotent POST must NOT be retried on an ambiguous 5xx

SOURCE · 2026-09-17

The dated verification record documents 1,526 passing tests with network access blocked, plus fresh-wheel installation checks.

REPORT · 2026-09-04

Independent snapshot provenance, responsibility decomposition and explicit equivalence limits.

SOURCE · 2026-09-17

Dated offline engineering validation and remaining real-account acceptance boundary.

SOURCE · 2026-09-17

Ambiguous non-idempotent creation is not automatically retried.

SOURCE · 2026-09-17

History includes breadth-first cross-sectional sweep; this is a related branch of membrane.

SOURCE · 2026-09-17

Historical membrane snapshot with paper-rig hardening history; group with parent.

SOURCE · 2026-09-17
CONTINUE IN QUANTITATIVE FINANCE & MARKETS

Statistical arbitrage & module discovery