Goodfire
An AI interpretability research lab building tools to decode neural networks and turn AI into something that can be understood, debugged, and intentionally shaped like software.
At a Glance
- AI Researchers
- Model Developers
- Enterprise AI Teams
- Scientific Research Organizations
AI Tools by Goodfire
(1)
Silico
AI Model Interpretability Platform
Discussions
No discussions yet
Be the first to start a discussion about Goodfire
Latest News
Goodfire Raises $150M at $1.25B Valuation to Design Models with Interpretability
Goodfire Named to 2026 CNBC Disruptor 50 List
Goodfire and Radical AI Partner to Advance AI-Driven Discovery in Materials Science
Announcing Our $50M Series A to Advance AI Interpretability
Products & Services
A hosted mechanistic interpretability API that allows developers to decode and control model internal behavior through feature steering and activation harvesting.
A developer toolkit for integrating model interpretability into production workflows, supporting SAE (Sparse Autoencoder) training.
Market Position
Goodfire is the first company to offer a hosted API for mechanistic interpretability, positioning itself as the 'debugging platform for Software 2.0'. It competes with in-house interpretability teams at frontier labs like Anthropic and OpenAI, as well as emerging AI safety and eval startups.
Leadership
Founders
Eric Ho
Co-Founder and CEO. Previously co-founded RippleMatch, an AI-driven recruitment platform. Early investor in AI safety and interpretability.
Tom McGrath
Chief Scientist. Previously a Research Scientist at Google DeepMind, where he was a founding member of their mechanistic interpretability efforts.
Dan Balsam
CTO. Previously VP of Engineering at RippleMatch and held engineering roles at several high-growth tech startups.
Executive Team
Eric Ho
CEO
Co-founder with a background in applied AI and scaling tech startups (RippleMatch).
Tom McGrath
Chief Scientist
Expert in neural network interpretability from Google DeepMind.
Board of Directors
Founding Story
CEO Eric Ho founded Goodfire after realizing that powerful AI systems were being deployed without anyone understanding how they actually work. The company's vision is to move beyond the 'black box' of AI toward 'intentional design,' where models can be inspected and fixed from the inside out.
Business Model
Revenue Model
Usage-based API business model, where revenue is generated through token consumption and subscription tiers for specialized access.
Pricing Tiers
Access to core interpretability features with standard rate limits.
High-volume access, dedicated support, and custom model integration for large-scale deployments.
Target Markets
- AI Researchers
- Model Developers
- Enterprise AI Teams
- Scientific Research Organizations
- AI Safety and Alignment
- Debugging model hallucinations or biases
- Fine-grained control over model personas
- Optimizing models for specific scientific domains (e.g., materials science)
- Rakuten
- Apollo Research
- Haize Labs
- Radical AI