Making AI Practical
Our Investment in Snorkel
Artificial Intelligence (AI) will continue to reinvent every sector in the economy over the coming years. Natural language processing can interpret and react to human language; computer vision lets machines interpret and act on the physical world; and intelligent decision engines will lead to more effective decisions and outcomes. At the core of this reinvention is leveraging data to deliver insights and predictions using AI.
Given the incredible potential of AI to drive business transformation, we’ve seen rapidly growing enterprise investment in AI infrastructure and solutions. Enterprises increased AI spending by 62% last year, with spending expected to reach nearly $100B in 2023. Despite this growing spend, most Fortune 500 CIOs we speak to are still disappointed by their investment, and aren’t making progress using AI as quickly and as effectively as they expected to.
Why is this the case? Current infrastructure efforts and investment have missed the key to enabling AI – the data. Architectural advancements and compute innovations can only take companies so far, if the heart of modern AI – the training data – remains a major hurdle. The challenge, and value unlock, lives in helping customers understand, label, augment, and manage the data that high quality AI leverages and gets built on.
Today, we’re thrilled to announce our investment in Snorkel AI, which is coming out of stealth and announcing Snorkel Flow, their end-to-end, data-first AI platform. Snorkel Flow solves the training data problem with a fundamentally new programmatic approach developed over years at the Stanford AI lab. We co-led the initial seed with our friends at GV in Spring 2019 and after seeing the company’s rapid progress and customer momentum, led the Series A in Fall 2019. I’m privileged to have joined the board.
World-Class Team, Strong Open Source Traction
The Snorkel co-founders – Alex Ratner, Chris Ré, Paroma Varma, Braden Hancock, and Henry Ehrenberg – are a group of exceptional computer science faculty, scientists and industry engineering leaders originally from the Stanford AI Lab. The Snorkel founding team is at the rare intersection of state of the art technical research, industry experience, and customer centricity and pragmatism.
At the Lab, the team spent five years creating the Snorkel Project – a new programmatic approach to labeling, building, and managing training datasets as the key to driving ML development. The team developed and deployed Snorkel’s core technology in partnership with many of the world’s leading organizations like Google, Intel, Apple, Stanford Hospital, the DoJ and many more. The core technical advances underpinning Snorkel have been featured in over thirty peer-reviewed publications.
As they saw the impact their technology was having on organizations’ AI efforts, the Snorkel team realized there was an opportunity to reimagine the entire machine learning development and deployment process. The team spun out of the Stanford AI Lab in 2019 to start Snorkel AI and bring Snorkel Flow to market.
Pioneering a New Data-First AI Paradigm With Snorkel Flow
Snorkel Flow is the data-first platform for enterprise AI, enabling customers to quickly build and iterate from unlabeled data sets to high quality ML deployed in production. Instead of relying on manual, hand-labeled data, users programmatically label and manage the data by writing “labeling functions” and other programmatic operators, which express rules or heuristics that make it straightforward for subject-matter experts to map their domain expertise to ML-based solutions. Snorkel Flow trains, deploys, analyzes and monitors ML models and workflows on top of this programmatic training data. Users can easily improve and adapt these models just by editing their programmatic training data. Snorkel Flow connects every step of the ML pipeline back to the piece that matters: the data.
Today, Snorkel Flow is used by a growing number of customers including large financials, government agencies, and other Fortune 500 enterprises. Snorkel’s customers are dramatically accelerating the pace at which they can get AI deployed, and leveraging Snorkel’s eyes-off approach to unlock new AI use cases with sensitive data. Snorkel Flow integrates with datasets and supports machine learning workloads across on-premise and hybrid public cloud environments.
Greylock has a distinguished history of partnering with market-defining companies from the earliest stages including Workday, Palo Alto Networks, and Sumo Logic, and more recently Abnormal Security, Utmost and Obsidian. We take great delight in rolling up our sleeves and helping founders build from the beginning. We’ve partnered with Snorkel AI since the company’s formation (they initially worked out of Greylock offices in Menlo Park) and have been closely supporting them through Seed, Series A and beyond. We’re thrilled to be partnered with Alex and team on this journey as Snorkel makes AI practical and leads the intelligent enterprise transformation.