Openlayer is the company behind one of the world’s most comprehensive platforms for testing AI. To learn more about Openlayer, Pulse 2.0 interviewed the company’s co-founders Rishab Ramanathan, Vikas Nair, and Gabriel Bayomi.
Background Of The Founders
The three founders met while working at Apple as ML engineers specializing in helping teams identify and fix failures in their machine-learning models. And over the course of working across 15 different teams at Apple, they helped ensure responsible AI through high-quality data and developed a robust evaluation process for machine learning models.
Formation Of Openlayer
While working at Apple, the company founders found that the validation and testing of ML models was one of the biggest challenges for engineers. And with the unpredictability of machine learning, the founders asked themselves: “How can we guarantee that a model works for every situation possible? How can we prevent bugs, underperforming cohorts, or biases before the model is rolled out?”
If machine learning was a game of whack-a-mole, then Openlayer aims to be the tape that covers up the holes and problems before they pop up. Plus applying guardrails to smoothly transition machine learning into the real world.
Core Products
What are Openlayer’s core products and features? The company founders pointed out that the core product of Openlayer is a machine learning validation and testing platform that helps teams detect, diagnose, and correct model failures before models enter production and monitor performance over time. Plus Openlayer helps teams systematically improve models and datasets by:
— Confirming the integrity of training and validation datasets
— Highlighting meaningful discrepancies between training, evaluation, and production data
— Ensures that models meet their target performance benchmarks
— Validates that the models are robust to edge cases by generating synthetic data for injecting noise and conducting adversarial attacks
— Assuring fairness of model behavior across data subpopulations
— Follows versions of models and datasets and compares their performance
— Legitimizing model behavior by surfacing which features of the data were used to make a prediction
Use cases:
Openlayer’s platform helps teams improve their models and datasets by verifying the integrity of training and validation. And this ensures that models meet their target performance benchmarks and guarantees the fairness of model behavior across all data subpopulations.
If problems and discrepancies were to arise, then Openlayer identifies and surfaces them between training and evaluations. And tactics like generating synthetic data to inject noise and conduct adversarial attacks to validate models to edge cases. Via probabilistic testing, teams are able to run tests on all grounds and ensure data integrity for their projects, fostering a smooth transition into the real-world environment.
Biggest Milestones
What have been some of Openlayer’s most significant milestones? And since the company’s last funding round, Openlayer is now able to support Large Language Models (LLM).
Funding
The seed funding round brings the total raised to $4.8 million in equity. And the funding will be used to enhance platform functionality and support additional machine-learning tasks. As of right now, our team consists of seven employees.
Plus the funding will enable the Openlayer team to create more sophisticated guardrails for customers to test their models against as they iterate. And the platform will also enable edge-case detection using synthetic data to generate test cases they might not have considered. Plus customers will benefit from faster and more organized development velocity.
Quiet Capital is known as the lead investor with participation from Y Combinator; Picus Capital; Hack VC; Liquid2 Ventures; Mantis VC; Jonathan Swanson, founder of Thumbtack and an investment consultant for Greystone Consulting; Mike Krieger, co-founder, Instagram; Max Mullen, co-founder, Instacart; Guillermo Rauch, CEO and founder of Vercel; Gokul Rajaram, member of the board of directors, Coinbase, Pinterest, and The Trade Desk; Immad Akhund, co-founder CEO, Mercury; Oliver Cameron, VP, Product, Cruise; Yuri Sagalov, managing partner, Wayfinder Ventures; Rodrigo Schmidt, head of engineering, NPE at Meta; and John Kim, CEO, SendBird.
Differentiation From The Competition
What differentiates Openlayer from its competition? Even though most competitors focus on monitoring models in production Openlayer sets itself apart by focusing on the evaluation of models even before deployment, guided by responsible AI principles according to the company founders.
Favorite Memory Working For Openlayer
What has been a favorite memory working for Openlayer? “Our favorite memory working for Openlayer has to be the moment when a user uncovers previously unconsidered data or model limitations. Witnessing the ‘eureka’ moment as they grasp the implications of these hidden issues is truly inspiring. It’s a powerful reminder of how our work at Openlayer sheds light on the unseen aspects of machine learning, enabling our clients to fine-tune their strategies and move forward with greater clarity and confidence,” said the company founders.
Challenges Faced Building The Company
What have been some of the challenges faced in building the company? “The industry faces challenges in differentiating their message in the crowded MLOps and Generative AI market. Openlayer sets itself apart by focusing on evaluation and responsible AI,” acknowledged Openlayer’s founders. “With the recent boom in generative AI, one of the biggest concerns in the industry is how these models can potentially misbehave, resulting in biased or inappropriate outputs. Openlayer addresses this concern by implementing stringent guardrails and monitoring mechanisms to ensure that their AI models are used responsibly and ethically, thus minimizing the risks associated with their deployment.”
Total Addressable Market (TAM)
What is the total addressable market (TAM) size that Openlayer is pursuing? The company founders analyzed the MLOps market being expected to be around $6 billion by 2027 and it is growing extremely fast. And the overall AI software market is estimated to be a $100 billion market. And with the new models introduced this year, the tendency is that it will likely outgrow expectations.
Future Company Goals
The company founders concluded that Large Language Models (LLMs) have shown remarkable potential in comprehending and generating human-like text, making their evaluation crucial for further refinement. And the primary objective is to become the go-to platform for establishing guardrails for LLMs. Openlayer assists engineers in identifying and addressing potential biases for any Machine Learning task, thus facilitating the deployment of high-performing, ethical, and responsible AI systems.
