At MoBagel, I own the architecture and implementation of AI agent workflow and AI knowledge-base backend services, along with an ML experiments-building platform. I scaled parallel knowledge-base file uploads by 8x by decoupling a memory-heavy module.
I also refactored data connector services and the ML processing pipeline, reducing idle workload to 20% and deployment load by 30%. I introduced OTel-based observability that reduced dataflow and log tracing time to 20%, and automated development environment setup with a code agent, reducing setup time by 95%.
At Aiello, I developed and maintained a chatbot system and implemented an NLP semantic search system that improved music playback accuracy by 10%; this work resulted in a patent. I also built a signal classification module at Insyde and created Career Polaris, a project that analyzes strengths and recommends roles based on evidence.
