LinkedIn(opens in a new tab)GitHub(opens in a new tab)
All projects

Data

Vaahan

Real-time hours-of-service compliance for trucking fleets

  • TypeScript
  • Node.js
  • AWS Lambda
  • Amazon SQS
  • Amazon SNS
  • AWS SAM
  • PostgreSQL
  • Drizzle ORM
  • React
  • GitHub Actions
  • Samsara API
  • Claude Code

Architecture (simplified)

ELD APISamsaraPollersLambdaevery 10minPostgreSQLRDSDashboardReactViteEval queueSQSRulesengine7 layersNotifiersSNSELD APISamsaraPollersLambdaevery 10minPostgreSQLRDSDashboardReactViteEval queueSQSRulesengine7 layersNotifiersSNS
Driver clocks are polled from the ELD API every ten minutes into PostgreSQL and an evaluation queue; a seven-layer rules engine writes violations back and publishes to SNS for the notifiers. Dead-letter queues with redrive back every queue.

About the project

Vaahan was a real-time hours-of-service compliance platform for trucking fleets. Fleets usually find out about a violation the next day, when the fine is already on its way. Vaahan watched every driver's hours-of-service clock around the clock, caught violations within minutes, and predicted breaches before they happened. I co-built it with a partner who runs a trucking compliance business: I architected the system and wrote its rules engine.

Under the hood it was an event-driven serverless pipeline on AWS: about twenty Lambda functions on SQS and SNS that polled the Samsara ELD API every ten minutes, with idempotent ingestion into PostgreSQL, per-carrier FIFO queues, and dead-letter queues with redrive. The rules engine was deterministic and written from scratch, seven layers implementing the FMCSA hours-of-service rules. Before cutover it ran in shadow against the old engine for 25 days: over 360,000 evaluations, about 99% agreement and no regressions in the cutover week. Then the old engine was retired.

It shipped through 175 automated deploys on GitHub Actions, and I built it with AI-driven development: a multi-agent Claude Code setup with 14 custom skills.