Suhaanth V V

Software Engineer

Silo Fortune Pvt Ltd

Aug 2025 - Present

Architected a four-stage asynchronous pipeline (Event-Driven Architecture) on Redis Streams (enrich, decide, deliver, schedule) serving two products across 10 languages, tuned to 100-message batches at 50 concurrent sends per worker and scaled horizontally by adding replicas to the same consumer group. Guaranteed at-least-once delivery through per-stage idempotency keys, a 30-second XPENDING/XCLAIM recovery sweep, and three-attempt exponential backoff on transient failures, so a worker that dies mid-message replays it rather than losing it. Moved event types, templates, A/B variants, and campaign audience SQL into PostgreSQL. The growth team now launches a campaign or rewrites copy without waiting on a deployment. Closed a duplicate-send risk across replicas by electing one scheduler leader per tick with a Redis lock, TTL heartbeat, and automatic failover, keeping campaign runs over audience segments of up to 150K users single-sourced. Instrumented all four workers with structured JSON logging (structlog) shipped to AWS CloudWatch Logs Insights, binding trace_id, event_id, and delivery_id through contextvars so a single notification stays queryable end-to-end across workers sharing one log stream. Tracked delivery outcomes as Redis counters flushed to PostgreSQL per campaign and language, and corrected reporting that had conflated device-token reach with true FCM delivery success, giving the growth team an honest delivery rate to plan against.