Galileo
To help AI teams move from monitoring AI failures to stopping them by providing a complete agent reliability platform that brings unit testing and CI/CD rigor to the AI development lifecycle, making AI trustworthy, safe, and observable.
Founding Story
Founded in 2021 by former employees from Google AI, Google Brain, and Uber AI who were fed up with high AI project failure rates and the amount of time data scientists spend on menial data preparation tasks. The founders realized that AI developers could only be successful if they could manage the unpredictable nature of LLM applications. They started Galileo to create a machine learning data tooling stack and a collaborative system of record for AI model development, pursuing a vision of AI reliability.
Discussions
No discussions yet
Be the first to start a discussion about Galileo
Leadership
Founders
Vikram Chatterji
CEO and Co-Founder. Previously worked at Google AI for three years as Product Management Lead, where his team built large-scale applications using language models including work on BERT models. Has led product building from 0 to 1 multiple times across geographies, company sizes and business models, used by 100M+ users globally.
Atindriyo Sanyal
CPO and Co-Founder. Previously led engineering at Uber AI (Michelangelo) where he was co-architect of Uber's feature store, and was a senior software engineer at Apple focusing on Siri, developing the earliest versions of Siri. Played key engineering roles in influential AI infrastructure built at Uber and Apple.
Yash Sheth
CTO and Co-Founder. Previously led the speech recognition platform team at Google as staff software engineer, managing the Google Speech Recognizer platform and growing speech recognition 800x. Spent nearly nine years at Google working on speech-recognition models that recognize spoken words from audio.
Executive Team
Vikram Chatterji
Co-founder and CEO
Former Google AI Product Management Lead who built large-scale applications using language models
Atindriyo Sanyal
Co-founder and CPO
Former Uber AI and Apple Siri engineering leader, co-architect of Uber's feature store
Business Model
Revenue Model
SaaS subscription model with tiered pricing based on API usage, traces/month, and enterprise features. Revenue streams include: (1) Free tier for developers and small teams, (2) Pro tier with monthly/annual subscriptions scaling with usage, (3) Enterprise tier with custom pricing for unlimited scale and on-premise deployments.
Pricing Tiers
For developers and small teams. Includes 5,000 traces per month, unlimited users, and unlimited custom evals.
Built for growing apps. Pricing scales based on number of traces. Includes everything in Free plus 50,000 traces per month, standard RBAC, advanced analytics & insights, and dedicated Slack support.
For teams needing unlimited scale and security. Includes everything in Pro plus unlimited traces, custom rate limits, deployment options (Hosted, VPC, or on-prem), enterprise-grade security (RBAC, SSO), dedicated CSM, real-time guardrails, 24/7 support, low-latency dedicated inference servers, and forward deployed engineering support.
Target Markets
- Enterprise AI teams
- Fortune 500 companies
- AI developers and data scientists
- Machine learning engineers
- GenAI application developers
- Technology companies
- RAG (Retrieval-Augmented Generation) systems
- AI Agents and agentic workflows
- Multi-agent AI systems
- Enterprise AI safety and security monitoring
- High-stakes environments requiring strict CI/CD for AI
- Rapid debugging and deployment of LLMs
- Comcast
- Twilio
- HP
- Magid
History & Milestones
Launched Agentic Evaluations framework and toolchain for AI agent reliability
Released Luna-2, next generation of small language models for real-time AI evaluations and guardrailing
Announced free Agent Reliability Platform for multi-agent AI systems
Launched Galileo Protect, a real-time hallucination firewall to safeguard enterprise generative AI
Released Galileo Luna, first-of-its-kind Evaluation Foundation Models (EFMs) for enterprise GenAI evaluations
