At OmniSwitch, I built a high-performance Go AI gateway that exposes 1,600+ models through a unified API. It includes five routing strategies, CEL-based routing, circuit breakers, shadow traffic, and atomic configuration hot-reload.
I also built gateways for A2A and MCP protocols, with agent discovery, task execution, streaming, and CEL-gated tool execution. To reduce provider round trips, request latency, and API costs, I implemented exact-match and semantic/vector caching using embedded SQLite.
For AI security, I built CEL-based guardrails to detect and block prompt injection, SQL injection, and PII. I also developed a hybrid Go/Rust-WASM guardrail engine that delivered 1.2–1.4× faster scanning, and shipped Python and Node.js SDKs.
At Walmart, I optimized Spark, Scala, and Snowflake processing to improve query performance by 5× and reduce operational costs by 50%. I also developed Python and Java microservices for inventory management, reducing API latency from 2.1 seconds to 800 milliseconds while maintaining 99.5% availability.
