At UrbanLayer, I built a parcel feasibility product that brings more than 25 public data sources into one profile, with a RAG copilot over the Chicago Municipal Code. I wrote a primary-source benchmark and raised copilot accuracy from 68% to 96%.
At The Official AI, I replaced one-off pollers with a shared multi-node PostgreSQL job queue for 15 AI task types and rebuilt the LoRA fine-tuning pipeline, protecting about $5K MRR across 22 customers. Earlier, I shipped APIs and patient-facing interfaces for Abraxis Networks’ telehealth platform and cut response latency by 66% on a Soul Machines feature.
