Jack Lee

A modular Python platform for researching, simulating, and paper-testing trading strategies with explicit risk controls.

• Separated market data, strategy, risk, execution, journaling, and analytics into testable components.
• Built time-aware research workflows with causal features, chronological evaluation, and bid/ask execution modelling to avoid misleading backtests.
• Orchestrated repeatable paper sessions with risk checks, trade journals, and automated verification so results can be inspected and reproduced.

• Python
• PyTorch & scikit-learn
• Time-series validation
• Causal feature engineering
• Risk & execution modelling

A developer tool that routes bounded engineering tasks by capability, risk, and cost while keeping human review in the workflow.

• Uses explicit task and provider policies to select a model and effort level without hard-coding one vendor.
• Shares routing, execution, and run telemetry across the CLI, interactive TUI, and MCP server.
• Supports isolated worktrees, ownership boundaries, and verification steps for reviewable code changes.

• TypeScript & Node.js
• MCP
• Policy-driven model routing
• Process orchestration
• Isolated worktrees

View on GitHub →

Turns timetable screenshots and assignment briefs into a realistic study schedule that fits around lectures, commitments, and deadlines.

• AI only extracts structure from messy inputs; workload estimation and scheduling stay deterministic and explainable.
• Workload is derived from ECTS, assessment weighting, and rubric structure, not guessed by the model.
• Preserves completed work when plans change, with validated local data and offline support in a PWA.

• Next.js & TypeScript
• Cloudflare Workers
• Multimodal extraction
• Deterministic scheduling
• Offline-capable PWA

Visit PlanAround →

Building and maintaining a US deal-discovery site with automated candidate research and human review before publication.

• Built searchable deal pages and item alerts around a structured catalogue.
• Developed a Python ingestion pipeline that checks product data, suppresses duplicates, and prepares candidates for review.
• Kept publishing behind manual price and listing checks, with automated site builds and deployment.

• Hugo
• Python
• GitHub Actions
• Product-data validation
• Review-gated publishing

Visit DealLedger →