I build AI agents for a living — and run one for myself, 24/7. CoFounder & Head of AI at Phot.AI. Ex-AppyHigh. Ironman 70.3.
Akshit Raja.
- 600M+app downloads shipped
- 55Mmonthly users at peak
- 70.3Ironman finisher, Tallinn
- 24/7personal agent uptime
built & worked with
MS · Northwestern University · BTech · PEC Chandigarh
~/about
Pick a goal. Build the loop.
Iterate.
I'm Akshit. I build things with AI and I build my own body — and I'm increasingly convinced those are the same skill: pick a goal, break it into something measurable, set up feedback loops, and keep iterating.
Today I'm CoFounder & Head of AI at Phot.AI, where we build agentic creative workflows for ecommerce — testing dozens of ad and listing variations in parallel so brands can find what actually sells. Before that I co-built AppyHigh, a mobile app studio that crossed 600M+ downloads and 55M monthly users.
My path wasn't a straight line: derivatives trader, analytics roles at Snapdeal, ixigo, and Adidas, a data-science master's at Northwestern, a mechanical engineering degree from PEC Chandigarh. The throughline is the same — learn fast, build fast, read the signal, double down or kill it.
Outside work I'm an Ironman 70.3 finisher (Tallinn) now training for Hyrox, and I'm a little obsessed with feeding my own training and recovery data into AI agents to coach myself.
Right now I'm building my personal AI agent — one that runs 24/7 on a tiny server, builds its own tools, and does real work before I wake up — and I'm documenting the whole thing in public. If that's your kind of thing, get on the newsletter.
~/now-building
Building my personal AI,
in public.
An agent that runs around the clock on a tiny server, builds its own tools, and does real work before I wake up. Every step documented — nothing polished away.
$ agent status
● online — doing real work while I sleep
03:14 analyzed 12 ad variants, killed 9
03:42 rebuilt its own scraper tool — v3
04:05 drafted newsletter issue from notes
05:58 pulled Garmin data, updated training plan
$
- The exact setupthe agents, the server, the tools — nothing hidden behind "it depends"
- The real costswhat this actually costs per month to run, in rupees and dollars
- What brokefailed runs, dead ends, and the fixes — the part everyone else edits out
- What workedworkflows you can copy the same week
~/in-public
Follow the build.
Short, real updates from the build — agent demos, costs, fails and wins.
~/beyond-the-screen
I train like I build —
measure everything, trust the process.
phot.ai, recognized by
- AWSGlobal Top 80
- Google for StartupsTop 16 India
- NVIDIAInception
- JioGenNext
- Product HuntProduct of the Day & Week