At Quralyst AI, I architected a multi-provider LLM orchestration pipeline with dynamic vendor switching, delivering about 8% cost savings and a 30% accuracy improvement. I also built a microservice that cut processing costs by 30%, improved result quality by 50%, and increased speed by 10%.
I designed a high-throughput backend for long-running asynchronous AI tasks and engineered MongoDB schemas for multi-stage, high-concurrency workflows. I deployed containerized services on AWS EC2 using Docker and ECR.
At Callify.ai, I shipped speech-to-text and text-to-speech services with AI-driven call summarization. I improved system efficiency by 1.5x and reduced costs by 7% through backend automation.
My live, self-built projects include Studio, an AI product suite with speech-to-text and a self-hosted Qwen 2.5 7B chat microservice, plus a centralized SSO service and a multi-tenant household finance platform. I also built a Discord bot for controlling local AI coding agents, with isolated sessions and guarded Git automation.
