Devashish Upadhyay | AI Engineer & Founder of Ziplo

My Story

At 24, I've already built AI systems that most engineers only read about. From the lakes of Udaipur to the tech scene of Sydney, I've been obsessed with one question: why do 90% of AI agents never make it to production?

I found out the hard way. At Greenstone, we built 70+ AI agents for a regulated insurer. Only 7 survived to production. Not because they weren't good—because testing them was nearly impossible. That failure rate? It's not unique to us. It's an industry-wide crisis.

So I'm fixing it. Ziplo is the testing infrastructure that will define how the next generation of AI gets deployed. The question isn't if this becomes the standard—it's who gets in early.

I jump out of planes for fun. Building startups is the easy part.

"The best time to invest in AI infrastructure was 2 years ago. The second best time is now."

What I'm good at

§ AI & Multi-Agent Systems

§ Python & Backend Architecture

§ Cloud Infrastructure & DevOps

§ Regulatory Compliance

§ System Design & Integration

What I'm Building

The projects keeping me up at night

ZIPLO

The infrastructure layer for trustworthy AI

Every company building AI agents faces the same bottleneck: testing. It's eating 70% of development time industry-wide. Ziplo eliminates that. We're building the testing and compliance infrastructure that lets teams ship AI agents 10x faster—without the risk.

Prompt injection detection. Hallucination catching. Compliance validation. Everything enterprises need to deploy AI with confidence. We're not just solving a problem—we're creating a new category.

3 Enterprise pilots live

70% Dev time lost to testing

90% AI agents fail to ship

Background

Greenstone

Where Ziplo was born

Built 70+ AI agents for a regulated insurer. Learned firsthand why 90% never make it to production—and decided to fix it.

AI Engineer II 2023-2025

Project

AI Voice Agents

1,000 calls per minute

Built real-time telephony AI handling outbound calls at scale. Twilio, Azure Speech, LiveKit, WebSockets—the full stack for voice AI.

Greenstone 2024

Project

AI Quality System

60+ agents auditing every call

Multi-agent system for automated QA compliance. Reduced manual auditing by 90% while improving accuracy.

Greenstone 2024

Tech Stack

What I Work With

Languages, tools, and platforms

Python, FastAPI, LangChain, OpenAI, Azure (Functions, Cosmos DB, Speech), Twilio, PostgreSQL, Redis, Docker, Terraform

AI · Backend · Cloud

Education

SP Jain Global

Bachelor of Data Science

Sydney, Australia. Built the foundation for everything that came after.

CGPA 3.3/4

Certifications

Microsoft Azure

AI · Data · Cloud

AI Engineer (AI-102), Data Scientist (DP-100), Data Analyst (DA-100)

3x Certified

Let's Talk

Open to Opportunities

Whether it's investment, collaboration, or something interesting I haven't thought of yet—I'm always open to a good conversation.

Get in Touch

Check out Ziplo

See what we're building to fix AI testing.

Ziplo.ai