RJ
Bengaluru, India

Rajamohan Jabbala

AI/ML innovation leader and founder of CLS++. Twenty years turning hard research problems into products people trust.

Founder, AlphaForge AI Labs · Ex-Amazon Alexa · Two-time CTO

0+
Years building AI
0M+
Users on platforms scaled
0
Patents filed
0
0→1 products founded

“An agent that can't trust its own memory can't be trusted with anything that matters. I'm building the reliability layer that earns that trust.”

Why I'm building CLS++

Agent memory layers kept silently dropping facts and blindly overwriting them — and you'd only find out in production. CLS++ is a memory engine that doesn't: it archives every superseded fact with full lineage, survives embedder failures, and lets you prove what changed. Memory you can actually trust.

The pedigree

Chief Technology Officer

Bonito Designs
  • Built Project Raven — cut design retrieval from 1 day to 500ms
  • Shipped AI sales agent “Suchi”, automating 400+ calls a day
  • Stood up MLOps that lifted scalability 40%

Chief Technology Officer

Convergent Inc
  • Led a $200M enterprise cloud transformation
  • Built real-time ML analytics pipelines
  • 92% reduction in reconciliation time

Software Development Manager

Amazon — Alexa
  • Launched the Alexa × Prime Video platform (3.5% conversion)
  • Created “Clairvoyant” — capacity planning from 1 week to 1 minute

M.Tech, Computer Engineering

IIT Madras · 1997–2000

M.S, Applied Statistics

Osmania University · 1995–1997

Patents & invention

Jul 2024

Cutlist Optimization Ensemble Algorithm

Govt. of India

2025

Thermodynamic memory architecture (CLS++)

Provisional

2014

Elastic Neighbourhood

Govt. of India

2000

Raster Snap function for Vectorization

IEEE

Selected work

CLS++

Memory engine

Reliable memory for AI agents — archives every superseded fact with full lineage, survives embedder failures, and lets you prove what changed.

Full lineageEmbedder-failure safe

Agent Transport Protocol

Rust · protocol

The TCP/IP of AI agents — a five-layer stack for trust-aware, economically-optimal multi-agent networking. ~37k lines of production Rust.

−53% cost28× context compression0 failures / 10k tasks

deepAudit

AI SaaS

AI-powered software deep audit — 750+ signals across 40 categories. One API call to audit any GitHub repository.

750+ signals40 categories

Quant Trading Engine

Python · fintech

A quant-based trading engine and service built to put institutional-grade strategy automation in the hands of retail traders.

Live tradingStrategy automation

Off the clock: finished the 300km Paris-Brest-Paris randonnée in under 15 hours. The same appetite for long, hard problems shows up in the code.

Build on memory you can trust

Give your agents a memory layer that never silently loses or wrongly overwrites a fact.