Hapax

Most AI tools hand a user a blank prompt and expect them to become the expert. We built Hapax to invert that relationship. The platform observes how a team actually works, maps the events that trigger real effort, and builds AI co-workers that handle the downstream response automatically, in hours and without any engineering time. It is not another chat interface or a search box over a team's files, but a proactive AI that learns how a business operates and builds itself around it.

I joined in 2024 as a senior engineer on a small team and have contributed across two full product generations, including the expansion from a strictly B2B product for financial institutions into a broader B2B and B2C surface. My day-to-day work is end-to-end on the LLM-powered product surfaces that customers actually touch: chat, agent workflows, and the tool-calling features that let an agent reach beyond the model to complete real tasks. I am consistently among the top contributors in shipped work and the trusted on-call engineer for production incidents and cross-team fixes.

Beyond feature work, I invest in the performance and reliability of the platform itself. The clearest example is a round of APM trace noise that had been quietly bottlenecking the application. I diagnosed the source, eliminated it, and dropped latency on key API and LLM calls from minutes to milliseconds. In a product deployed inside more than twenty financial institutions, from $250M credit unions to $90B banks, where the quality bar is unforgiving, that kind of change is what separates software that feels heavy from software that feels like it gets out of your way.