hello world!
logo

Accelerate ML Teams

Empower your ML teams to deliver robust and performant AI models by auto-diagnosing and improving your data and models.
Book a demo
Proud to work with leading organizations
SBB CFF FFSUS Army

Our Platform

LatticeFlow Platform

The only platform that can auto-diagnose data and models, empowering ML teams deliver robust and performant AI models faster.

Smart City
Medical
Manufacturing
Mobility
CASE STUDY

Smart City Analytics

Univrses used LatticeFlow to create performant AI models within few days, by auto-diagnosing the model and improving it with synthetic data, saving months of manual work for data curation and labeling.
Latticeflow Platform Infosec Pipeline

<2 weeks

Time to adapt the AI model to a new domain
Saving months of labeling and engineering effort per deployment.

10+

Blind spots found 
and fixed
Observed in new deployments.

up to 3x

mAP improvement*
*On deployments in new domains.
CASE STUDY

Major Medical Company

A major medical company used LatticeFlow to analyze its datasets and AI models, discovering previously unknown blind spots and spurious correlations in a fully secure on-premise deployment.
flow-normal

10+

Critical blind spots found and fixed
Discovered in state-of-the-art models deployed at hospitals.

<15 min

Time to discover first issues
Rapidly and interactively discover actionable insights into improving data and models.
CASE STUDY

Quality Inspection

In the challenging task of automated defect detection with highly imbalanced datasets, LatticeFlow auto-diagnosed & fixed inconsistent labels and doubled model robustness.

89%

Precision of auto-diagnosing and fixing inconsistent labels
Applied to challenging datasets annotated by domain experts.

3 hours

From installation to custom data quality rules
Saving months of manual work.

+38%

Robustness improvement
Without sacrificing any model accuracy.
CASE STUDY

Safety Assessment of Traffic Sign Detectors

The German Federal Office for Information Security used LatticeFlow to assess the safety and robustness of state-of-the-art traffic sign detectors.

30+

Robustness specifications defined
Covering camera noise, sign stickers, shadows, and others.

10+

Unknown failure modes uncovered
Confirmed with real-world images on which the model systematically fails.

70%

Certified Robustness
While improving model accuracy by 0.2%.

Our Mission

Robust and trustworthy AI

Our mission is to change the way the next generation of AI systems are built. If we are to use AI in our businesses, at doctor’s offices, on our roads, or in our homes, we need to build AI systems that companies and users can trust.

AI leadership

We are building the foundations of robust and trustworthy AI. Our co-founder Prof. Krause is founding chairman of the ETH AI Center and co-directs the Swiss Data Science Center. Prof. Vechev’s group has built the world’s best verifier for deep neural networks and teaches the largest public course on Reliable and Trustworthy AI.

#1 verifier for deep neural networks

Our robustness verifier achieved the highest score at VNN-COMP'20, the first competition for certifying neural networks, outperforming teams from MIT, Oxford, UIUC, and others.

OUR EXPERTISE

Cutting-edge research applied in practice

We are leading AI professors and researchers from ETH Zurich with broad expertise in formal methods, symbolic reasoning, and machine learning. We started LatticeFlow with the goal of building the world’s first platform that enables companies to deliver robust AI models that work reliably in the wild.

Top research

We have published 50+ peer-reviewed publications in the area of robust and trustworthy AI at top-tier venues in machine learning, programming languages and security, including ICML, NeurIPS, PLDI, POPL, and S&P.

Entrepreneurship

We have experience in bringing research into production and have previously built two successful ETH spin-offs acquired by PwC and the unicorn Snyk.

Leadership team

We bring expertise from the world's best universities
Petar Tsankov
Co-founder and CEO
Petar holds a PhD from ETH Zurich. Before LatticeFlow, he co-founded ChainSecurity which was acquired by PwC. He graduated from Georgia Tech with two research awards and was awarded by the President of Bulgaria for his achievements in computer science.
Pavol Bielik
Co-founder and CTO
Pavol holds a PhD from ETH Zurich, where he worked at the intersection of machine learning and programming languages. For his work, Pavol received a highly competitive Facebook Fellowship, as the only student from Europe, for 2017 and the Romberg Grant in 2016.
Martin Vechev
Martin Vechev
Co-founder
Martin is an Associate Professor at ETH Zurich. He holds a PhD from the University of Cambridge and was a research staff member at the IBM T.J. Watson Research Center. Previously, Martin co-founded two ETH spin-offs which were acquired by PwC and the unicorn Snyk.
Andreas Krause
Andreas Krause
Co-founder
Andreas is a Professor at ETH Zurich. He holds a PhD from CMU and was an Assistant Professor at Caltech. He is founding chairman of the ETH AI Center and co-directs the Swiss Data Science Center. He received Test of Time awards at ICML 2020 and KDD 2019.

Investors and advisors

Andreas Goeldi
Andreas is a partner at the btov Partners. Internet technologist, entrepreneur, investor with 25 years of experience. He co-founded and built successful startups in Europe/US, and now invests in AI/ML, devops, martech, and B2B SaaS.
Benedikt Körling
Benedikt is a partner at GFC, a billion-dollar venture fund with 16 offices around the globe investing across all stages with a focus on early-stage investments. Benedikt is heading GFC's Berlin office and invests in B2B SaaS, DevTech and Gaming in DACH, CEE, and Benelux.
Boris Paskalev
Boris is a technologist and business entrepreneur who has co-founded and built successful companies in Europe/US. He was part of the disruptive growth and IPO of the unicorn Vistaprint, and in 2020 he exited his startup DeepCode to the unicorn Snyk.

News

LatticeFlow demos its robust AI platform at ECR 2022

Join us from July 13th until July 17th at the ECR 2022 in Vienna, the largest annual conference in Europe ...
Read More

LatticeFlow wins Swiss AI Award 2022

The Swiss AI Award honors the most promising and innovative AI start-ups from across Switzerland. The award was funded by ...
Read More
2022 CB Insights AI100

LatticeFlow Named to the 2022 CB Insights AI 100 List of Most Innovative Artificial Intelligence Startups

LatticeFlow, the platform for delivering robust and trustworthy AI, has been named to the prestigious CB Insights AI 100 List ...
Read More
crossmenu