If you would like an introduction to the Scale team, reply to this email.
Software Search: Scale
Company Snapshot
Founded: 2016
Employees: 600
Funding: $600M
Valuation: $7.3B
Stage: Series E
Locations: San Francisco, NYC, Remote
Company Overview
Scale accelerates the development of artificial intelligence.
Tell Me More
Founded in 2016, Scale has quickly become a household name in artificial intelligence and machine learning (AI/ML). Built off the belief that data is the primary building block to AI/ML, Scale initially launched with an unassuming product offering — data labeling. The company got its start by supplying autonomous vehicle companies with the requisite labeled data needed to train ML models to develop and deploy robo-taxis, self-driving trucks, and automated bots used in warehouses. As the key ingredient to training any ML model, data labeling was the perfect wedge to helping Scale execute its goal of bringing AI to all companies, regardless of technical know-how.
What’s Data Labeling Again?
At its core, data labeling is the process of identifying raw images and tagging them with meaningful text labels so that an ML model can process and learn from them. These images could be pictures of stop signs which enable autonomous vehicles models to learn when to stop, uploaded receipts so an expense reimbursement system can verify your corporate reimbursement, or even your tagged Facebook photos so Facebook can automatically tag you the next time your grandma uploads a family photo.
Having labeled data is the key ingredient to building any ML workflow, but historically, data labeling has been a huge bottleneck for in-house AI/ML teams. Depending on the use case, ML models could require hundreds of billions of data types to build performant models. This is no easy task and is truly mission-critical, especially when you consider the scale of these projects
Before Scale, in-house ML engineers would have to label data themselves or outsource it to offshore third-party data labeling companies. However these off-shore providers usually struggle to provide accurately labeled data delivered in a quick time frame, which is critical for the data’s usability. Furthermore, utilizing an in-house ML engineering team full of Phds to label data themselves is not cost effective and takes time away from higher-level engineering activities.
Enter Scale. Scale’s value proposition was simple to understand and allowed teams to get back to building and deploying models, not labeling data. With Scale, teams can outsource data labeling to the company’s in-house team of data labelers or leverage Scale’s automated data labeling tools to do it themselves. The company’s suite of data labeling offerings allow ML organizations to work with a reputable data labeling provider that can offer fast, efficient, and secure data labeling services.
With data labeling at its core, Scale has grown to 600+ employees, $200M+ in ARR, and is operating in multiple geographic regions. Today, Scale is rapidly expanding beyond data labeling to satisfy the end-to-end needs of ML practitioners.
Market Opportunity
AI/ML has often been touted as the next breakthrough industry. There are countless articles that have been written by analysts like McKinsey and Gartner that talk about the operational benefits gained from implementing AI/ML into business practices. Un-coincidentally, underpinning the entire AI/ML workflow is clean, usable data. Every ML model requires labeled data to instruct the model on what to learn, predict, or suggest. To put ML’s reliance on data into perspective, Open AI’s pre-trained text-to-text GPT-3 model was trained on 175 billion pieces of labeled data, and that’s just one type of model! That’s not even including the classic big-tech names who are all constantly running ML models and algorithms to serve up the best music recommendations via Spotify or DTC brands on Instagram.
Scale’s ability to service the immediate pain point of ML engineers has given it the agency to build for the long-tail of the market. It is my belief that AI/ML has largely failed to deliver on the promise of its value to everyday businesses. Over the past decade, there’s been enormous progress in advancing the capabilities of AI/ML, driven primarily by new model architectures and the decreased cost of compute. However, integrating models with real-world business processes is still a technically challenging, lengthy, and expensive endeavor for most companies. Additionally, ML talent has largely been siloed to elite institutions, severely handicapping the everyday business’ ability to leverage bespoke AI/ML solutions.
While a handful of companies are able to systematically build, leverage, and deploy ML models to customers, the long tail of the market opportunity comes from operationalizing ML for the everyday company, or the “Have Nots”.
Today, companies like Scale are building horizontal applications and infrastructure to service the entire AI/ML market beyond just data labeling for elite institutions. Over the past decade, a heap of startups have been founded to tackle all aspects of the AI/ML workflow, including model monitoring (e.g., Arthur AI, Arize), point solutions for legacy industries (e.g., Hyperscience, ScienceIO), synthetic data creation (e.g., Synthesis AI, Gretel AI), and more.
When analysts talk about the long tail value of the ML market, they’re talking about enabling all industries, regardless of technical talent, to leverage sophisticated models to achieve desired business outcomes.
Why I Like Scale
With data labeling at its core, the company has grown to 600+ employees, $200M+ in ARR, and is operating in multiple geographic regions. Today, Scale is rapidly expanding beyond data labeling to satisfy the end-to-end needs of ML practitioners.
Scale’s initial data labeling offering proved to be a strong wedge into the burgeoning AI/ML market, helping the company reach $100M ARR in less than 4 years.
As the company continues to scale its data labeling offering, it’s simultaneously diversifying its product line to service the full life cycle of AI development across teams. Today, Scale is a multi-product organization building and deploying scalable ML infrastructure, point solutions, and even synthetically generated data to various buyers across the enterprise. These new product offerings from data generation to model performance to data extraction, and beyond, combined with Scale’s flagship data labeling offering position it to service the entire end-to-end AI/ML value chain.
Data labeling for ML teams was initially a nonobvious wedge into the enterprise, but the lion share of their market opportunity is in larger contract value, higher margin SaaS products like model monitoring software and point solutions like their newly released, Document AI.
Having a multi-product portfolio allows Scale’s sales team to “land and expand” — sell data labeling services to AI/ML teams, and as the sales rep learns more about the customer’s AI/ML needs, sales reps can suggest additional products and gain warm introductions to higher contract value sales. Eventually, I believe Scale’s go-to-market team will be able to segment the entire market based on the “haves” vs. “have nots” of AI/ML talent. Given Scale’s reputation in AI/ML, I believe these land and expand sales will be much easier vs traditional cold outreach many startups and incumbents are leveraging today.
Additionally, there are other ways for the company to grow as well. Scale could acquire Hugging Face and become the de facto marketplace for pre-trained models, the company could begin selling data to practitioners for more ethical AI, or they could become a data platform themselves. Given the vast amount of data flowing through the Scale platform, it’s possible that they could leverage their customer data by redacting personal identification information and sanitizing it with synthetically generated data to build their own pre-trained models for various industries.
Lastly, beyond Scale’s revenue growth or market opportunity, it is clear that the company is composed of insanely talented, hard working people. Scale is filled with ex-founders and FAANG-level product managers and engineers. When I spoke with the team and asked about their future goals, nearly everyone I chatted with mentioned being a founder themselves, or leading product lines.
In the blink of an eye, Scale could be the next big thing.
Scale’s Hiring Corner
Similar Companies
Reminder
As always, you can find the abbreviated list of companies we’ve already talked about here. Below are links to the previous posts