Johnson James Kanjirathinkal
I'm Johnson. At present, I'm building Skillfully, a public benefit corporation in SF, where I spend most days wrestling with multi-agent systems and LLM evals. This site is where I log a few things about me.
Work
At Skillfully we run live simulations, voice, adaptive chat, tool use; for enterprise teams. We built a modular multi-agent framework to keep long, multi-turn conversations from falling apart.
Before Skillfully, I worked on enterprise ML across Asia, on multilingual text analytics and customer-experience platforms. alongside roles. I've spent time at Lawrence Berkeley National Laboratory, Geckolyst and various startups. MEng in Industrial Engineering & Operations Research from UC Berkeley with a research focus on ML optimization and IP analytics.
UC Berkeley Startup Innovation Award. GSV Cup 50. I judge at Berkeley StEP and Big Ideas, and the occasional contest in between
Writing
Essays on Substack.
- Why I keep three tabs open
The industry keeps trying to delete the interface. I keep three tabs open for a reason.
- Nobody demos a blank box
Why open-ended prompts reward users who arrive knowing what to ask. Most people don't.
- The judge gets flattered first
What grading co-authored transcripts revealed about reward signals in LLM judges.
Elsewhere
Contact
I invest in and advise a few early-stage AI companies. Say hi: