Software Engineer at Optmyzr
Building production AI and full-stack systems for PPC marketing: Sidekick, a conversational AI agent, a multi-agent natural-language-to-DSL rule builder, a RAG support bot over internal knowledge, and a portfolio management platform aggregating ad performance data across major ad networks.
- Built agent orchestration, tool-access harnesses, prompt flows, guardrails, and eval benchmarks for reliable tool calling across internal optimization services.
- Built a multi-agent system that turns plain-English requests into executable DSL programs for the company's Rule Engine, with a Pinecone retrieval harness and a compile-and-repair loop that feeds compiler errors back to the agent for self-correction — improving accuracy from 35% to 82% and cutting latency from 200s+ to 45-60s.
- Created a RAG support system over 10,000+ help articles, conversations, internal threads, and product documents with refreshed embeddings and citation-oriented answers.
- Developed portfolio tooling over 1,000+ portfolios and 3,000+ accounts, with GraphQL, React, MongoDB, DuckDB, .NET, and C#.
Tech stack: C#, .NET, JavaScript, React, GraphQL, MongoDB, DuckDB, Redis, AWS, AI agents, RAG, Pinecone