LatticeFlow AI is developing the AI Assessments platform that integrates governance frameworks with the technical validations required for regulatory compliance, including the EU AI Act. The company positions the platform as the world’s first to seamlessly combine governance and technical evaluation for AI systems. LatticeFlow emphasizes closing the gap between governance processes and deep technical validation so enterprises can deploy reliable, compliant models. The project is framed to address the growing AI Trust, Risk, and Security Management (AI TRiSM) market, which the article cites as projected to exceed USD 8 billion by 2033. LatticeFlow has collaborated on COMPL-AI, a compliance-centered evaluation framework for LLMs developed with ETH Zurich and INSAIT. The company is also working with KPMG Switzerland in a strategic partnership to accelerate AI risk assessments and help enterprises meet emerging regulations. Recent external funding will be used to advance the platform and its commercial opportunities. LatticeFlow, an ETH Zurich spin-out from 2020, provides tooling that automatically diagnoses issues in computer vision models and improves both data and models. The product inspects pre-trained vision models and the underlying datasets to find labeling issues, poor-quality samples and model blind spots. Customers use the platform to reveal hidden correlations and failure scenarios that standard data checks miss. The company offers its solution as an on-prem tool today and says it may offer a fully managed service in the future. LatticeFlow counts more than 10 customers in Europe and the U.S., including Siemens and Swiss Federal Railways, and is running additional pilots. The team will use new funding to hire aggressively to service customers and expand its product portfolio. LatticeFlow is an AI startup spun out of ETH Zurich in 2020 that builds tools to help teams develop and deploy safe, reliable, and trustworthy models. Its platform provides systematic ways to monitor models, assess reliability across relevant data slices, and supply tools for improving model robustness beyond aggregate metrics like accuracy. The founding team includes CTO Pavol Bielik, CEO Petar Tsankov, and ETH professors Martin Vechev and Andreas Krause, who bring expertise in verification and adaptive learning. LatticeFlow is currently in early access and already has pilot customers including Swiss railways (SBB), Germany’s Federal Cyber Security Bureau and the U.S. Army, and is working with other large enterprises on computer vision models. The company raised a $2.8M seed round led by btov and Global Founders Capital to accelerate product development and acquire new customers. The team plans to build out a U.S. presence in the near future.

Total raised

$18M

Funding rounds

3

Latest round

Equity

Latest activity

Oct 2024