Technology

Microbrain Technology is built as a two-layer system: private trading infrastructure for execution and data processing, and a public-facing publication site for transparent documentation and performance summaries.

Private layer (secure)

Market data ingestion, feature engineering, model inference, and order execution. Secrets and brokerage credentials never leave your private environment.

Public layer (read-only)

Next.js on Vercel for publishing research notes, risk disclosures, and performance snapshots via read-only endpoints.

Stack overview

Execution

Interactive Brokers (IBKR) via IB API / Gateway

Research

Python (feature engineering, backtests, model training)

Data

Minute bars, bid/ask, engineered technical features, event filters

Deployment

Vercel (Next.js) for publication + lightweight APIs

Ops

Scheduled runs, logs, alerting, and report generation

Security

Secrets kept off Vercel; read-only publishing endpoints

Engineering principles

Separation of concerns

Trading execution stays on your private infrastructure; the website is a read-only publication layer.

Reproducibility

Clear assumptions, versioned code, and deterministic report outputs where possible.

Risk-first design

Position sizing, stops, exposure limits, and kill-switches are part of the system—not afterthoughts.

Leak-free research

Time-aligned features, realistic execution assumptions, and strict separation of train/validation/test regimes.

Next integration

Next we can add a read-only endpoint (e.g. /api/performance) that serves your daily summary JSON, and render it on the Performance page.