Live product / public reference

Prediction Market Operations

A live multi-user operating system for discovering sports markets, routing strategy decisions, executing through guarded wallets, and reconciling order state against on-chain activity.

Illustration of a control desk with three lanes of sliders and toggles — IllustrationAI-generated illustration of the operations control plane. It shows no real or synthetic market data.

At a glance

Period: 2026; active private product and public reference
Role: Product and systems work for live operations
Team: Private side-project team; separate public reference implemented by Franco
Users: Operators managing automated sports-market workflows across multiple user accounts
Status: Private live operation; separate public reference implementation

ProductLive multi-user system
ModesPaper and production
ExecutionGuarded per strategy
Public proofBounded reference repo

My part

Worked across sports-market discovery, per-user strategy routing, wallet onboarding, guarded CLOB execution, order-state reconciliation, dashboard and Telegram operations, tests, and release diagnosis.

Tradeoff

Show the operational product and its engineering boundaries without exposing strategies, credentials, users, wallets, production data, infrastructure, or financial claims.

Evidence

The private product operates live. The linked repository is a separate paper-only reference with clean provenance, bounded public reads, idempotent state, diagnostics, tests, containers, and browser acceptance.

Live sport becomes an actionable market

The discovery layer combines live scoreboard data with Gamma metadata and CLOB liquidity. It resolves event identity, timing, market availability, and tradable outcome tokens before a strategy receives an opportunity.

Illustration of a ball turning into a ticket — IllustrationAI-generated illustration: a live event resolves into a tradable market.

Every user gets a controlled execution path

Configuration is independent by user and strategy. Each decision passes through its selected paper or production mode, isolated wallet, balance validation, kill switch, limits, and pre-trade checks before it can reach the CLOB.

Illustration of three separate lanes, each with its own gate and wallet — IllustrationAI-generated illustration: each user and strategy runs in its own guarded lane.

An order is not a result

Submission begins an order lifecycle; it does not settle the outcome. The system tracks acknowledgements, open quantity, partial and complete fills, cancellations, rejects, and active exits while preserving the intent that created the order.

Illustration of a chain of links that starts as a paper slip and ends in a check mark — IllustrationAI-generated illustration: an order stays connected across every lifecycle state.

Operations continue after execution

Exchange responses, recorded fills, and on-chain activity can disagree or arrive at different times. Reconciliation workers compare those surfaces, repair stale state idempotently, and expose exceptions through the dashboard and Telegram for diagnosis.

Illustration of two ledger blocks interlocking under a magnifying lens — IllustrationAI-generated illustration: reconciliation joins exchange and chain events.

Public evidence without exposing private logic

The linked Ops Lab is an independent, deliberately paper-only implementation. It demonstrates exact read allowlists, an idempotent ledger, diagnostics, tests, containers, and clean provenance without publishing private code, strategies, data, wallets, users, or infrastructure.

Illustration of a glass case with an engine model next to a closed box — IllustrationAI-generated illustration: a public reference beside private product logic.

Project linksGitHub