Website · 2026
Loaderhouse
loaderhouse.com · Architect and lead engineer
A freight load board for mid-size Indian carriers — public marketplace, broker console and a matching service that ranks loads against a truck's route, capacity and return leg.
p95 search latency, down from 11s
active carriers on the platform
loads matched in the last year
The problem
Loaderhouse was running on a PHP monolith where every search hit the primary database. Peak-hour searches took eleven seconds, brokers were losing loads to the phone, and the team could not ship a mobile client on top of it.
What I did
- Rebuilt search as a read-optimised projection in Postgres with Redis-backed geo indexes, taking p95 search latency from 11s to 240ms.
- Split the matching engine into its own service with a scored ranking model over route overlap, capacity and historical acceptance.
- Introduced an event log so the broker console, notifications and analytics all read the same stream instead of polling.
- Shipped a public marketing site and SEO-indexable load pages that now bring in a third of new carrier signups.
Architecture
Next.js front end on CloudFront, Spring Boot services on ECS Fargate behind an ALB, Postgres with read replicas, Redis for geo and session, and an SNS/SQS fan-out for notification and analytics consumers. Terraform for everything; deploys are trunk-based through GitHub Actions.
Request path
Similar problem on your side?