All selected work

Optimization / Full-stack systems / Project notes

RailBlockAI

Making room for railway maintenance. A planning prototype that prioritizes maintenance tasks and fits them into available timetable windows.

React.NETPythonScikit-learn
RailBlockAISystem overview
01React planner
02.NET API
03Python optimizer
04Schedule + review
Resource-aware schedulingArchitecture diagram

The problem

Maintenance work competes for limited timetable windows and shared resources. This prototype explores how a planner can review priorities, inspect a schedule, and approve or reject proposed blocks.

How the system works

A React dashboard displays a Gantt timeline and work queue. A .NET API serves a frozen replay bundle to a Python FastAPI optimizer, which ranks tasks and places them into available windows.

A key engineering decision

The optimizer combines prioritization with resource-capacity checks and cross-day coupling. A heuristic fallback is available when the trained prioritization model cannot be loaded.

Implemented scope

Schedule generation, triage, daily/weekly/monthly planning views, and approve/reject decisions are connected through a React dashboard and API. Demo decisions use in-memory storage.

Scope & limitations

The included data is a small synthetic fixture for a single corridor, not live railway operations. Weekly and monthly views replay a saved timetable snapshot; decisions reset on restart.

Next project

ShakespeareGPT

Explore case study