At SoftServe, I took ownership of Hexagon’s mining dashboard after the previous vendor, reverse-engineered undocumented parts of the system, and created a knowledge base to support its maintenance. I also proposed and developed an AI Harness for Spec-Driven Development.
On the Keller Williams high-load trading platform, I defined a migration from Elasticsearch to Mongo Atlas Search that reduced indexing and text-search costs by roughly 50%. I also co-implemented federated micro-frontends, helping improve load performance from red to green.
As Developer Team Lead on Keller Williams, I helped turn around an underperforming team and increased its velocity by 40%. I also improved service authentication and Kubernetes topology using SPIRE and Istio.
For Custom GPT Multimodal, I designed an asynchronous backend architecture and a prompt-versioning system, and abstracted the model layer to support OpenAI and Claude. Across these projects, I’ve worked on data engineering, full-stack development, architecture, and production AI integrations.
