Announcing our Seed Round
An old programming book suggests that engineers carry with them, at all times, a plastic rubber ducky. When stuck on a problem, the engineer is to immediately pull out the rubber ducky and begin explaining the situation. As wisdom has it, working with the rubber ducky will exponentially speed up finding clarity on the solution.
Your recommender system can make a lot of money, but behind the scenes, machine learning teams are slowed by managing bug fixes, feature requests, model iteration, experiments / A/B tests, and requests for data. Rubber Ducky Labs was built to be a trusted rubber ducky — a suite of tools to speed up machine learning teams, and return them to projects that have a direct impact revenue and engagement.
Rubber Ducky Labs builds operational analytics for recommender systems. We’re engineers with a background in retail tech, ML tooling, and infrastructure, having worked for Polyvore, SigOpt, Meta and Datadog.
Today, I’m excited to announce (TechCrunch) our $1.5 million seed investment round led by Bain Capital Ventures with participation from Cadenza Ventures, Y Combinator, and angel investors Brad Klingenberg (ex-Chief Algorithms Officer at Stitch Fix), Patrick Hayes (co-founder of SigOpt), and Dave Aronchick (co-founder of Bacalhau and Expanso). This funding allows us to expand our team and continue building and scaling our product.
Operational Analytics for Recommender Systems
At large companies, a 1% lift in recommender systems (RecSys) performance means a $1M+ lift in revenue, creating excitement around learning and experimentation. However, RecSys teams are often burdened with a large backlog of bug fixes, feature requests, model iteration, experiments, A/B tests, and requests for data. If it takes more than a few hours to answer a question like, “Why is our homepage recommending that I buy ski jackets in June?”, then ML engineers will move on to another task.
"Recommender systems are revenue drivers, but they don't fit well into existing ML infrastructure. We need better streamlined tools for ML in this space."
— James Kirk, former recommender systems staff ML engineer at Spotify
The solution is tooling that establishes a paved path for analyzing and iterating on recommender systems. Unfortunately, most MLOps tools are designed for generic ML use cases, so companies must build RecSys workflow tooling in-house. However, even when companies manage to launch tooling internally they encounter hidden maintenance burdens – from maintaining frontend code to maintaining team morale — that eventually lead to orphaned tools and burnt out engineers.
“Managing operations for recommender systems cost us one ML engineer’s worth of engineering time, but two ML engineer’s worth of morale”
— John McDonnell, former Director of ML (Recommender Systems) at Stitch Fix
At Rubber Ducky Labs, we build the visually-forward tools that your team needs to debug why your homepage is recommending ski jackets in June, and so much more. Our goal is to help machine learning teams learn, experiment, and iterate faster, so that they can get back to projects that directly impact revenue and engagement.
Become a Design Partner
We are seeking partners who share our vision for revolutionizing the way that recommender systems are developed and optimized. Contact us to learn more, and join our design partner program.
