Daniel Balsam

Daniel Balsam is an American engineer and the co-founder and chief technology officer of Goodfire, the San Francisco mechanistic-interpretability company he established in June 2024 with Eric Ho and Tom McGrath. The company develops Ember, a hosted interpretability platform that decodes the internal computations of large neural networks and exposes them as programmable features. As of May 2026, Balsam leads engineering and field-research at Goodfire after a $150 million Series B in February 2026 at a $1.25 billion valuation, following six years at RippleMatch where he was the founding engineer alongside Ho and rose to head of AI.

At a glance

• Education: Bachelor's degree, New York University, 2014; undergraduate research assistant at Columbia University's Department of Psychiatry from 2011 to 2012, with co-authorship on optogenetics work published in Molecular Psychiatry in 2014.
• Current role: Co-founder and chief technology officer of Goodfire, since June 2024.
• Key contributions: Founding engineer at RippleMatch (2018) and head of AI through 2024; co-author of "Understanding and Steering Llama 3 with Sparse Autoencoders" (Goodfire Research, December 2024); co-author of "Genome Modeling and Design Across All Domains of Life with Evo 2" (Nature, 2026); co-founder and chief technology officer of Goodfire (June 2024); architect of the Ember platform engineering and the field-research function that deploys interpretability tooling with enterprise customers.
• Public profile: Multiple appearances on the Cognitive Revolution podcast hosted by Nathan Labenz across 2024 to 2026; keynote on sparse autoencoders alongside Neel Nanda at a November 2024 Goodfire interpretability event.
• X / Twitter: @DanJBalsam
• LinkedIn: dbalsam
• Google Scholar: Dan Balsam

Origins

Balsam is American and attended New York University for an undergraduate degree completed in 2014. He has described his study as centered on neuroscience rather than computer science, telling Nathan Labenz on the Cognitive Revolution in August 2024 that "I studied neuroscience before I became an engineer working at startups." Public-profile aggregators including RocketReach describe the program as an Independent Study rather than a standard departmental major.

Concurrent with the NYU degree, Balsam worked from 2011 to 2012 as an undergraduate research assistant at Columbia University's Department of Psychiatry, producing a co-authorship on the 2014 Molecular Psychiatry paper "Optogenetic Stimulation of DAergic VTA Neurons Increases Aggression" with Catia M. Teixeira, Darshini Mahadevia, Jay A. Gingrich, and Mark S. Ansorge of the Sackler Institute for Developmental Psychobiology. His Google Scholar profile lists the optogenetics paper alongside a second 2014 Molecular Psychiatry contribution on dopamine and serotonin signaling in developmental periods.

Career

Balsam's first software-engineering role was at Controlco, a building-controls and energy-management company, from 2014 to 2016. He moved in 2016 to TixTrack, an events-ticketing platform in New York, as a software engineer through 2018.

In 2018 Balsam joined RippleMatch as the second engineer hired by the company, two years after Eric Ho and Andrew Myers had founded the early-career recruiting platform out of Yale. The Lightspeed Venture Partners profile of Goodfire and the Work-Bench Series A announcement describe him as RippleMatch's founding engineer, a positioning consistent with his role as the first engineer Ho recruited. Balsam moved through engineering-lead and engineering-manager roles before becoming head of AI in 2022, with the Lightspeed profile crediting him with "[building] out the core engineering organization and [deploying] LLMs in production." RippleMatch raised approximately $79.9 million during his tenure, including the June 2022 $45 million Series B led by Goldman Sachs Asset Management.

Balsam left RippleMatch in 2023 alongside Ho. The Contrary Research dossier describes the transition as a sabbatical during which he studied AI research full-time. Balsam has said the period crystallized into a decision that "there was nothing else that was worth working on besides minimizing AI risk." He reconnected with Ho over the same conviction.

Goodfire was incorporated in June 2024 with Ho as chief executive officer, Balsam as chief technology officer, and Tom McGrath, former founder of the Google DeepMind mechanistic-interpretability team, as chief scientist. McGrath had connected with Ho and Balsam in early 2024 after circulating a document at South Park Commons that mirrored their thesis. Balsam took responsibility for engineering, the Ember platform stack, and the field-research operation deploying interpretability tools with enterprise customers across language, biology, and other domains.

The fundraising trajectory has been compressed. The August 2024 $7 million seed was led by Lightspeed Venture Partners. The April 2025 $50 million Series A was led by Menlo Ventures with Anthropic participating, in what coverage characterized as Anthropic's first equity investment in another startup, and coincided with the public launch of Ember. The February 2026 $150 million Series B at a $1.25 billion valuation was led by B Capital, bringing cumulative funding to approximately $209 million.

Affiliations

• Controlco: Software engineer and engineering team lead, 2014 to 2016.
• TixTrack: Software engineer, 2016 to 2018.
• RippleMatch: Founding engineer through head of AI, 2018 to 2024.
• Goodfire: Co-founder and Chief Technology Officer, June 2024 to present.

Notable contributions

Balsam's record runs from early neuroscience co-authorships at Columbia through engineering leadership at RippleMatch into the Goodfire research and product output.

• "Optogenetic Stimulation of DAergic VTA Neurons Increases Aggression" (Molecular Psychiatry, 2014). Co-authored undergraduate paper with Teixeira, Mahadevia, Gingrich, and Ansorge of Columbia's Department of Psychiatry.
• RippleMatch engineering build-out (2018 to 2024). Founding engineer through head of AI across six years; built the core engineering organization through approximately $79.9 million in cumulative venture capital and a customer base including SAP, General Mills, eBay, and AB InBev. Deployed large language models into the production recruiting product.
• Goodfire co-founding (June 2024). Established the mechanistic-interpretability company with Eric Ho and Tom McGrath; took responsibility for engineering and field-research. Recruited Chris Olah of Anthropic as advisor.
• "Understanding and Steering Llama 3 with Sparse Autoencoders" (Goodfire Research, December 2024). Co-authored research preview with McGrath, Myra Deng, and Ho applying sparse-autoencoder feature decomposition to Llama 3.3 70B and Llama 3.1 8B. The technical foundation for the Ember platform launched in April 2025.
• "Genome Modeling and Design Across All Domains of Life with Evo 2" (Nature, 2026). Multi-author Arc Institute and Goodfire paper with Balsam among contributing authors. The most-cited paper in his Google Scholar profile.
• "Interpreting Evo 2" (Goodfire Research). Co-authored Goodfire research applying sparse-autoencoder feature decomposition to layer 26 of Evo 2 and exposing biological concepts in the model's internal representations.
• Public-talk record. Recurring appearances on the Nathan Labenz Cognitive Revolution podcast with Tom McGrath: the August 2024 launch episode "Popular Mechanistic Interpretability," the May 2025 "Philosophy, Practice & Progress" episode, and the March 2026 post-Series B "Don't Fight Backprop" episode introducing the Intentional Design research pillar; keynote on sparse autoencoders alongside Neel Nanda at a November 2024 Goodfire event.

Investments and boards

• Goodfire (AI): Co-founder and Chief Technology Officer, June 2024 to present. Privately held mechanistic-interpretability company; approximately $209 million in cumulative funding through three rounds, with the $150 million Series B in February 2026 at a $1.25 billion valuation.

No public personal angel-investor activity on record outside the Goodfire founder role in AI, semiconductors, datacenters, software, or energy as of May 2026.

Network

Balsam's longest-running professional relationship is with Eric Ho, his Goodfire chief-executive co-founder, with whom he has worked continuously since joining RippleMatch as founding engineer in 2018, through the AI-safety pivot of 2023 to 2024, and into the Goodfire founding. The continuous Ho-Balsam working relationship spans approximately eight years across two companies. The third co-founder, Tom McGrath, joined Balsam and Ho in early 2024 from Google DeepMind via South Park Commons.

The most distinctive external relationship is the advisory role of Chris Olah, the Anthropic co-founder who leads the company's interpretability research program. The Anthropic Series A investment, the first equity investment in another startup that Anthropic has disclosed, deepened the institutional relationship beyond the personal-advisor link. The broader interpretability community in which Balsam is a reference point includes Neel Nanda, the independent researcher cited alongside him in the November 2024 Goodfire keynote, and the DeepMind interpretability network from which McGrath joined.

The customer-and-collaborator network includes Patrick Hsu, co-founder of the Arc Institute and the Evo 2 DNA foundation-model team, cited in the Series A announcement as an early Ember collaborator. The Evo 2 work produced both the Goodfire research write-up and Balsam's contribution to the Nature 2026 paper. Other public Goodfire customers include the Mayo Clinic and Rakuten.

Position in the field

As of May 2026, Balsam is the chief technology officer of the only commercial mechanistic-interpretability laboratory to have crossed the unicorn-valuation threshold. The structural distinction in his profile is the combination of the six-year RippleMatch engineering-leadership credential, the founding-engineer-to-CTO arc with the same chief executive across two companies, and the field-research function he runs at Goodfire.

The credential profile is unusual relative to peer chief technology officers in commercial AI research. Balsam does not hold a doctoral degree, and his transition into AI research came from an applied-engineering rather than an academic-research background. The undergraduate-research period at Columbia and the resulting Molecular Psychiatry co-authorships are the only formal academic-research line on the public record before the Goodfire founding. The Lightspeed profile characterizes Goodfire's founding team as an operating chief executive paired with academic-research credentials in McGrath and Olah, with Balsam as the engineering anchor that connects research output to deployable enterprise product. The prior collaboration with Ho is the structural feature that distinguishes the Goodfire founding team from the more typical co-founder pairings in mechanistic-interpretability research.

Outlook

Open questions over the next 6 to 18 months:

• Ember adoption. Whether the platform's customer base scales materially across AI laboratories, regulated-industry enterprises, and biology-and-genomics research groups, and whether the field-research function continues to produce case studies beyond the Arc Institute Evo 2 collaboration.
• Engineering-organization scale-up. Whether the post-Series B engineering organization scales without diluting the research-engineering culture of the first eighteen months.
• Research-engineering balance. Whether Balsam continues to produce co-authorships at the cadence of the Llama 3 sparse-autoencoder preview and the Evo 2 collaboration, or whether commercial scaling compresses the published-research cadence.
• Public-communication cadence. Whether Balsam's interview record expands beyond Cognitive Revolution into conference keynotes or peer-reviewed publications.
• Founding-team stability. Whether the Ho, Balsam, and McGrath team remains intact through the post-Series B scaling phase.
• Field-research playbook. Whether the Goodfire field-research function becomes a documented playbook for translating interpretability research into deployed enterprise tooling.

Sources

• Daniel Balsam on LinkedIn. Public profile listing the Goodfire chief-technology-officer role and the prior RippleMatch tenure.
• Dan Balsam on X. Public X account, "Cofounder and CTO @goodfireai," joined December 2024.
• Dan Balsam on Google Scholar. Publication and citation record covering the 2014 Molecular Psychiatry co-authorships and the 2024 to 2026 Goodfire and Arc Institute publications.
• Optogenetic Stimulation of DAergic VTA Neurons Increases Aggression. The 2014 Molecular Psychiatry paper from the undergraduate-research period at Columbia University's Department of Psychiatry.
• Goodfire Raises $7M to Break Open the Black Box of Generative AI Models. The August 2024 seed-round announcement.
• Goodfire: Building Interpretable AI. Lightspeed Venture Partners profile describing Balsam as RippleMatch's founding engineer and head of AI.
• Announcing Our Investment in Goodfire. Work-Bench Series A perspective citing the Ho-Balsam working relationship from RippleMatch through Goodfire.
• Announcing Our $50M Series A to Advance AI Interpretability Research. The April 2025 Goodfire blog announcement of the Series A round.
• Understanding and Steering Llama 3 with Sparse Autoencoders. The December 2024 Goodfire research preview, with Balsam among the four named authors.
• Interpreting Evo 2: Arc Institute's Next-Generation Genomic Foundation Model. The Goodfire research write-up of the Evo 2 sparse-autoencoder analysis with the Arc Institute.
• Genome Modeling and Design Across All Domains of Life with Evo 2. The 2026 Evo 2 release with the public model and code, accompanying the Nature paper for which Balsam is among the contributing authors.
• Understanding, Learning From, and Designing AI: Our Series B. The February 2026 Goodfire blog announcement of the Series B round at a $1.25 billion valuation.
• Popular Mechanistic Interpretability: Goodfire Lights the Way to AI Safety. The August 2024 Cognitive Revolution episode covering the Goodfire founding and Balsam's neuroscience-to-engineering background.
• Mechanistic Interpretability: Philosophy, Practice & Progress with Goodfire's Daniel & Tom. The May 2025 Cognitive Revolution methodology episode with Balsam and McGrath.
• Don't Fight Backprop: Goodfire's Vision for Intentional Design. The March 2026 Cognitive Revolution post-Series B episode introducing the Intentional Design research pillar.
• Goodfire Business Breakdown & Founding Story. Contrary Research dossier covering the Ho-Balsam working relationship, the 2023 sabbatical, and the AI-safety pivot.