2:07 a.m.

Somewhere, someone is still reading.

Page 1,412. Of 2,000.

Kannappan Suresh

I build AI that reads everything, and shows its work.

Senior Software Engineer, AI/ML at ValGenesis

Before that: PayPal, an AI start-up, Kissflow and Saama · building software since 2021 · Chennai

The problem

For years, software tried to help with rules.

Rules break the moment a page looks different.

The problem

Then AI learned to read. Fast. Anything.

But sometimes it makes things up. With medicine or money, that's dangerous.

So how do you trust a machine that reads?

1 · It understands

It reads the page the way a person would.

2 · It shows its work

Every answer points back to the page and line it came from.

No source, no answer.

3 · It knows when to ask

When it isn't sure, it says so, and lets a person decide.

4 · It proves it

Every change is tested against known answers.

That's how I took a document system from ~70% to ~90% accurate.

6:04 p.m.

Same desk. Same 2,000 pages.

The reviewer only had to check the three it flagged.

The machine does the reading.People make the call.

And everyone goes home on time.

Written and engineered by

Kannappan Suresh

ValGenesis
Senior Software Engineer, AI/ML · 2026 – Document AI: accuracy ~70% → ~90%
Enegma
Founding team, AI start-up · 2025 – 2026 AI replies ~5× faster · new AI agents in ~1 day, not weeks
PayPal
Software Engineer 2 · 2024 – 2025 AI mapped the code: Venmo's system rebuild ~30% faster
Kissflow
Software Engineer · 2024 60M+ records moved, zero lost
Saama Technologies
Software Engineer · 2021 – 2023 APIs 20% faster

Tools

Python · LangGraph · RAG · Evals · FastAPI · PostgreSQL · Kubernetes · GCP · Azure · AWS

Education

B.Tech, Chemical Engineering, Anna University

Awards

4× Shining Star, Saama Technologies
INSPIRE Science Camp, Govt. of India (top 200)

Filmed in Chennai, India.

Let's build something people can trust.