Saketh Maddamsetty

I’m a Software Development Engineer II focused on designing end-to-end systems that turn messy, real-world inputs into reliable outcomes—especially in AI extraction, analytics, and production reliability.

At Coinbase, I owned the end-to-end design of a multi-model AI workflow for financial due diligence, extracting signals from unstructured documents to score clients against internal regulatory guidelines. I also engineered a Human-in-the-Loop feedback loop to resolve extraction conflicts and improved accuracy by 15%, while slashing manual review latency by 95% and reducing turnaround time from 1–2 weeks to 1–2 days.

I build for scale and uptime: I architected real-time analytics pipelines that sustain 100K+ events/sec with Kafka to ClickHouse ETL, led cross-team incident resolution in ClickHouse deployments to reduce recurring alerts by 90%, and implemented a configurable Snowflake → ClickHouse snapshot pipeline that scales to ~100+ tables and ~1M rows/day. Earlier, I delivered zero-downtime production migrations (saving $400K+ annually), created storage and caching services to cut load and latency, and published MicroBlind, a flexible file system middleware that outperforms NFS/Virtio-fs/9pfs by 1.5x–4x.