Zerotree LabsKanpur, India — Est. 2026
AI Research Lab

Intelligence,
without the noise.

We build AI systems that operate in real environments — for real people, at real scale.

4Production systems
1,000+Hours of media processed
6Indian languages supported
Scroll to explore
Worldview

Most AI moves fast and breaks things.
We move carefully and build things that last.

The real frontier of AI isn't about what models can do in a sandbox. It's about what systems can sustain under real load, with real data, for real users.

India has 1.4 billion reasons to build AI differently — different languages, different infrastructure, different economic realities. We don't adapt global AI for India. We build AI that starts from here.

Practical intelligenceover hype.
Systems that sustainover demos that impress.
Real deploymentover continuous prototyping.

India-first problem solving — Kanpur, U.P.

What we build
Language Intelligence01

Transia

Language shouldn't be a barrier to learning.

1,000+ hours processed · 6 Indian languages

Transia converts educational video into knowledge that anyone can access — regardless of language or medium. Multi-language transcription, synthesis, and text-to-speech built for the scale of Indian content.

We've processed over 1,000 hours of content across Hindi, Tamil, Kannada, Punjabi, Marathi, and Assamese. This isn't translation. It's knowledge infrastructure.

Learning Intelligence02

Froga

A thinking companion, not a chatbot.

Interactive · Framework-selectable · Gamified

MBA case studies are learned by doing, not by reading. Froga is an AI practice partner that puts you in the analyst's seat — asking the right questions, pushing your reasoning, and adapting to how you think.

Choose your framework: profitability, strategy, or rigorous multi-axis analysis. Then think harder. Froga won't give you the answer — it'll help you find it.

Business Intelligence03

Intellisap

Built for where India's economy actually runs.

RAG architecture · WhatsApp-native · MSME-focused

Most small businesses in India run on WhatsApp. Intellisap meets them there — an AI assistant built on retrieval-augmented architecture that handles customer queries, automates conversations, and surfaces business knowledge instantly.

Not a bot. Not a script. A system that understands context, retrieves accurately, and grows with the business. Real economic impact for the businesses that power India's economy.

Discovery Intelligence04

Sairch

Search should understand intent, not just keywords.

Multi-constraint · Semantic · E-commerce native

E-commerce search is broken. Users know exactly what they want — their language doesn't map to how products are catalogued. Sairch is a contextual discovery engine that handles multi-constraint queries the way humans actually think.

Ask for a gold necklace under ₹5,000 for a summer wedding in a minimalist style — and get exactly that. Semantic understanding layered over any product catalogue. The intelligence layer that discovery was missing.

Engineering philosophy

We work in production.
Not in notebooks.

The gap between a working demo and a deployed system is where most AI fails. We build for that gap — designing for edge cases, real data volumes, and production failure modes from day one.

We've shipped forecasting engines, semantic video intelligence, and enterprise chatbot platforms. All measured. All in production. All serving real users.

01

RAG Architectures

Retrieval-augmented generation designed for accuracy over recall — not just fast retrieval, but contextually faithful answers from the right knowledge sources.

02

Agentic Workflows

Systems that take actions, not just answers. We build agents that plan, execute, and recover — operating autonomously on tasks that would otherwise require human intervention.

03

Production ML

We don't stop at prototypes. Every model we train ships with inference infrastructure, monitoring, and the failure handling that separates a notebook from a deployed system.

04

Semantic Understanding

Transformer architectures tuned for specific domains — video, text, queries, conversations. Language models that understand the problem, not just the vocabulary.

“Every system we deploy is built to handle real load, real data, and real failure modes. We design for the edge case, not the demo.”

The person behind it

Prakhar
Sharma

FounderMachine Learning Engineer8+ years in production AI
Transformer architecturesRetrieval-augmented generationAgentic AI systemsSemantic video intelligenceEnterprise chatbot platformsScalable ML deploymentForecasting enginesLLM fine-tuning

Prakhar has spent eight years building AI systems that get deployed — not shelved.

His work spans transformer architectures, retrieval-augmented generation, and agentic AI design. He's led enterprise chatbot platforms, built semantic video intelligence systems, and designed forecasting engines — all deployed on cloud infrastructure, all producing measurable business outcomes.

Based in Kanpur, he believes the most interesting AI problems are the ones that haven't been solved yet — the ones with messy data, constrained infrastructure, and users who can't afford for the system to fail.

He's not building for the demo. He's building for what comes after.

“Translating complex business problems into intelligent software systems — that's the job. Everything else is commentary.”

Where we are

From Kanpur, for India.

We're based in Kanpur, Uttar Pradesh — a city with deep industrial roots and a growing technical community. We don't treat our location as a constraint. We treat it as context.

The problems we solve are closer to us here. The users we build for are neighbors. The infrastructure realities we navigate are real. AI built from here carries a different kind of intelligence — one that understands what it's actually like to operate in India.

CityKanpur
StateUttar Pradesh
CountryIndia
Time ZoneIST (UTC +5:30)
Collaboration

Let's build something that matters.

We work with people who have real problems and the patience to build real solutions. If that sounds like you, we want to hear from you.

ResearchersWorking on language, vision, or reasoning systems
FoundersWith real problems that need real solutions
EngineersWho think in systems and build for production
OrganizationsLooking to deploy AI with measurable impact
or find us on LinkedIn

Response within 48 hours · No pitch decks required