Ashr
Ashr helps developers ship AI agents with confidence by mimicking production environments and users to catch regressions and fix failures before users find them.
At a Glance
- AI Startups
- Enterprise AI Engineering Teams
- Developers building LLM agents
AI Tools by Ashr
(1)Ashr
AI Agent Eval Platform
Discussions
No discussions yet
Be the first to start a discussion about Ashr
Latest News
ashr Joins Y Combinator's W26 Batch
Ashr: Mimic Your Production Environment and Users to Catch Agent Fails - Launch Post
Published 'Why's This Shit Broken? Our pivot story and why AI testing infrastructure is fundamentally broken'
Products & Services
A modular SDK-based testing and evaluation platform for AI agents that mimics production environments and user patterns.
Market Position
Ashr differentiates itself by focusing on 'mimicking real users' and 'production environments' rather than just static evaluations, providing a more robust testing framework for complex agent behaviors.
Leadership
Founders
Rohan Kulkarni
Co-founder at Ashr. EECS student at UC Berkeley. Previously built and exited an AI-native survey platform for old-age homes.
Shreyas Kaps
Co-founder at Ashr. UC Berkeley student. Former AI-native workflow engineer at Tessera Labs, mentored by ex-Meta and Netflix researchers. Worked on ERP transformations for Fortune 100 companies.
Executive Team
Rohan Kulkarni
Co-founder
EECS @ UC Berkeley, previous exit in AI survey tech.
Shreyas Kaps
Co-founder
UC Berkeley, former AI engineer at Tessera Labs.
Board of Directors
Founding Story
Started by Berkeley students Shreyas and Rohan while building AI automation for small businesses. They realized current testing tools couldn't handle AI-native edge cases, leading them to pivot to build the infrastructure for testing AI agents.
Business Model
Revenue Model
B2B Subscription / Usage-based for AI testing infrastructure
Pricing Tiers
Custom pricing based on evaluation needs
Target Markets
- AI Startups
- Enterprise AI Engineering Teams
- Developers building LLM agents
- Catching agent regressions
- Validating agent response quality
- Testing complex tool call paths
- Simulating production edge cases
- UC Berkeley
- Stanford University
- Human Behavior
- Pax Historia