Sooth Labs

Sooth Labs is an American artificial intelligence lab founded in 2026 by former Meta vice president Yaser Sheikh, former Meta innovation-program director Chuck Hoover, and Carnegie Mellon University professor Russ Salakhutdinov. The company is headquartered in Pittsburgh and develops AI models that produce probability estimates for geopolitical and market events, intended for customers in finance, defense, insurance, and real estate. As of early 2026, Sooth Labs had raised $50 million at a $335 million post-money valuation, led by Felicis Ventures with angel investments from Yann LeCun and Jeff Dean.

At a glance

• Founded: 2026 in Pittsburgh, Pennsylvania.
• Status: Private. $335 million post-money valuation as of the seed round.
• Funding: $50 million seed round at a $335 million post-money valuation. Led by Felicis Ventures, with angel investments from Yann LeCun (Meta AI / FAIR chief AI scientist) and Jeff Dean (Google chief scientist). Andrew Bosworth (Meta chief technology officer) serves as advisor.
• CEO: Yaser Sheikh (co-founder, formerly vice president at Meta and visiting professor at Carnegie Mellon).
• Other notable leadership: Russ Salakhutdinov (co-founder, professor at Carnegie Mellon, former first director of AI research at Apple, chief scientist of Magic), Chuck Hoover (co-founder, formerly directed billion-dollar innovation programs at Meta).
• Open weights: None disclosed.
• Flagship products: No public product release as of April 2026. Research focus on probabilistic forecasting models for geopolitical and market events.

Origins

Sooth Labs was founded in early 2026 in Pittsburgh by Yaser Sheikh, Chuck Hoover, and Russ Salakhutdinov. The founding team combines former Meta executives with Carnegie Mellon University faculty connections.

Sheikh had been a vice president at Meta before founding Sooth Labs. He continues to hold a visiting-professor appointment at Carnegie Mellon University's Robotics Institute, where his research has spanned computer vision, machine perception, and human-pose estimation. Hoover directed billion-dollar innovation programs at Meta during his tenure there. Salakhutdinov is the UPMC Professor of Computer Science in the Machine Learning Department at Carnegie Mellon, the first director of AI research at Apple from 2016 through 2020, and chief scientist of Magic, the San Francisco coding-foundation-model lab. He earned his PhD as a doctoral student of Geoffrey Hinton at the University of Toronto, with co-authorship on multiple foundational deep-learning papers.

The Pittsburgh headquarters location is unusual within the broader cohort of well-funded 2025-to-2026 AI Insurgents, which are concentrated overwhelmingly in San Francisco. The location reflects the founding team's strong Carnegie Mellon University ties and the broader Pittsburgh AI ecosystem, in which Carnegie Mellon University's School of Computer Science is a leading research center.

The seed round of $50 million at a $335 million post-money valuation, led by Felicis Ventures, was announced in early 2026. Angel investments came from Yann LeCun (Meta's chief AI scientist) and Jeff Dean (Google's chief scientist), both senior figures across the broader AI research community. Andrew Bosworth, Meta's chief technology officer, joined as an advisor.

Mission and strategy

Sooth Labs's stated mission is to train AI models that predict the likelihood of geopolitical and market events. The technical premise is that large-scale probabilistic forecasting can be approached as a foundation-model problem: AI systems trained on cross-industry datasets covering historical events, news, market data, and adjacent signals can produce calibrated probability estimates for future events that human analysts and traditional statistical models cannot match.

The strategy combines three threads. First, foundational research on probabilistic-forecasting AI architectures, building on the founders' research backgrounds in Bayesian methods, deep generative models, and probabilistic deep learning. Second, multimodal model design that supports inputs across video, audio, and text, allowing users to query the probability of specific future events directly through natural-language questions. Third, vertical specialization on customer segments where event-probability estimates have direct commercial value: finance (trading and risk management), defense (geopolitical-risk assessment), insurance (loss-event probability), and real estate (location-and-development-risk assessment).

The competitive premise is that probabilistic forecasting is structurally underaddressed by current frontier large-language models, which are trained primarily for next-token prediction rather than calibrated probability estimation, and that a vertically specialized lab can build defensible commercial positions across the four target customer segments.

In demonstrations published with the seed-round announcement, the Sooth model produced specific probability estimates for events such as a 33% probability of Anthropic going public and a 16% chance of a World Health Organization pandemic declaration by 2028. The framing positions Sooth Labs's product as directly comparable to prediction-market and forecasting-platform offerings, but with foundation-model scaling rather than human-aggregation as the underlying mechanism.

Models and products

• Foundational research output. Sooth Labs's primary public output as of April 2026 consists of the founding-team announcement, the company's stated technical thesis, and the demonstration probability estimates published with the seed-round announcement.
• No shipped models, APIs, or open weights. The company has not released a model, an API, or open weights as of April 2026.

The commercial distribution strategy beyond research output has not been comprehensively disclosed. The four named customer segments (finance, defense, insurance, real estate) imply a sales-led enterprise distribution model rather than a self-serve API or consumer product surface.

Benchmarks and standing

Sooth Labs has not released a model and is not represented on the standardized capability leaderboards as of April 2026. The standardized leaderboards measure language-and-reasoning capability rather than probabilistic-forecasting calibration, and may not be the appropriate evaluation framework for Sooth Labs's research output.

The closest existing benchmarks are within the academic forecasting literature (Brier scores and adjacent calibration metrics on prediction-market data) and the prediction-market industry itself. Direct evaluation of Sooth Labs's research output against frontier models will likely require new benchmarks specific to event-probability calibration at scale.

The company's standing rests on the founders' research credentials at Meta, Apple, Carnegie Mellon, and the academic ML community, the Felicis Ventures lead, the Yann LeCun and Jeff Dean angel-investor signals, and the Andrew Bosworth advisor relationship.

Leadership

As of April 2026, Sooth Labs's named leadership consists of:

• Yaser Sheikh, co-founder and chief executive. Formerly vice president at Meta. Visiting professor at Carnegie Mellon University's Robotics Institute. Public face for the company on technical claims and product framing.
• Russ Salakhutdinov, co-founder. Professor of computer science at Carnegie Mellon. First director of AI research at Apple from 2016 through 2020. Chief scientist of Magic. Doctoral student of Geoffrey Hinton at the University of Toronto. Co-author of foundational deep-learning papers including "Reducing the Dimensionality of Data with Neural Networks" (Hinton and Salakhutdinov, Science 2006).
• Chuck Hoover, co-founder. Formerly directed billion-dollar innovation programs at Meta.

Andrew Bosworth, Meta's chief technology officer, serves as an advisor. The senior research and engineering team beyond the named founders has not been comprehensively disclosed.

Funding and backers

Sooth Labs's funding history through April 2026 consists of a single closed round: the $50 million seed round at a $335 million post-money valuation. The round was led by Felicis Ventures.

Angel investments came from Yann LeCun (Meta's chief AI scientist and one of the three deep-learning Turing Award co-recipients) and Jeff Dean (Google's chief scientist and a foundational figure in Google's AI infrastructure history). Andrew Bosworth, Meta's chief technology officer, joined as an advisor.

The combination of senior-academic-and-industry angel investors and a single venture-capital lead is unusual within the 2025-to-2026 Insurgent cohort and reflects the founders' deep ties across both Meta and the broader academic ML community. The company has not disclosed cumulative compute commitments, cloud-partner relationships, or follow-on financing plans.

The reported valuation of $335 million is consistent with mid-size 2026-vintage Insurgent labs founded by senior-team founder cohorts but is differentiated within the cohort by the specialized probabilistic-forecasting positioning.

Industry position

Sooth Labs occupies a structurally distinctive position within the 2026-vintage Insurgent cohort. The combination of senior Meta executive and Carnegie Mellon faculty founders, the probabilistic-forecasting specialization, the Pittsburgh headquarters, and the angel-investor list (Yann LeCun and Jeff Dean) produces a profile not directly mirrored at any other lab.

The closest peer comparators are research-first Insurgents pursuing vertical specialization within specific application domains. Mirendil targets scientific reasoning in biology and materials science. Periodic Labs targets adjacent scientific applications. Hippocratic AI targets healthcare applications. The probabilistic-forecasting positioning is more commercially differentiated, with parallel competition from prediction-market platforms and traditional quantitative-finance forecasting providers rather than other AI Insurgents.

The strategic risks are substantial. The company has not released a model, the probabilistic-forecasting thesis has not been validated through shipped product or independent calibration evaluation, and the four named customer segments (finance, defense, insurance, real estate) span sales motions that often require dedicated go-to-market teams. The valuation depends on team credentials and angel-investor signals rather than capability evidence.

The strategic strengths are distinctive. Salakhutdinov's research credentials in Bayesian methods and probabilistic deep learning are among the deepest in the field. Sheikh's vice-president tenure at Meta indicates senior operating experience at scale. The Carnegie Mellon University connection provides talent access through one of the leading academic ML research centers. The Yann LeCun and Jeff Dean angel-investor signals provide credibility within the broader AI research community.

Competitive landscape

Sooth Labs competes with several Frontier, Insurgent, and adjacent vertical-AI labs and product-layer companies:

• Frontier labs (OpenAI, Anthropic, Google DeepMind). General-purpose frontier models are increasingly used for forecasting tasks through prompt-and-tool-calling patterns. Sooth Labs's specialized approach is positioned as architecturally differentiated.
• Mirendil and Periodic Labs. Vertical-specialized AI Insurgents with adjacent scientific-domain focus. Compete on the broader vertical-AI thesis but not directly on probabilistic forecasting.
• Prediction-market platforms (Kalshi, Polymarket, Manifold). Consumer and institutional-tier prediction markets that produce event-probability estimates through human aggregation. The closest existing product comparators on the forecasting surface.
• Quantitative-finance forecasting providers. Traditional financial-industry providers of event-probability and risk-assessment data. Compete on the finance-segment customer base.
• Academic forecasting research community. Carnegie Mellon University, where Salakhutdinov holds his appointment, is a particular point of overlap. The broader academic forecasting community produces some of the same calibration research that Sooth Labs targets.
• Defense-focused AI providers (Palantir, Anduril, Shield AI). Compete on the defense-segment customer base, though through different product surfaces.

Outlook

Several open questions affect Sooth Labs's trajectory in 2026 and 2027:

• The first published research output, including the company's specific probabilistic-forecasting architecture and any independent calibration evaluation.
• The first product release, including the target customer segment and product surface (API, application, or platform).
• The commercial strategy across the four named customer segments, which span dissimilar sales motions.
• Whether Sooth Labs accepts follow-on capital at a higher valuation, and on what timeline.
• Senior-talent recruitment from Meta, Carnegie Mellon, and the broader academic ML community.
• Strategic-partner relationships, particularly with finance-industry data providers and defense-industry contractors that could become channel partners or first customers.

Sources

• Gate News: Former Meta Executives Launch Sooth Labs, AI Event Prediction Startup Raises $50M at $335M Valuation. Primary source on the seed round and founding team.
• KuCoin: Sooth Labs, founded by former Meta executives, raises $50 million at a $335 million valuation. Funding-round coverage.
• From Samples to Scenarios: A New Paradigm for Probabilistic Forecasting (arXiv preprint). Adjacent academic research relevant to the company's technical thesis.
• Russ Salakhutdinov on nextomoro. Co-founder reference profile.
• Yaser Sheikh at Carnegie Mellon University. Visiting-professor reference page.