🚘 Uber Continues its Open-Source ML Traction

Weekly news digest curated by the industry insiders

Jan 16

📝 Editorial

When we think about active contributors to open-source machine learning (ML), we immediately gravitate towards big tech platforms providers like Google, Facebook, and Microsoft. We do not immediately associate companies like Uber with open-source ML contributions. However, the transportation giant has quietly become one of the most active sources of innovation for open-source ML projects. In the last few years, Uber has open-sourced over a dozen of ML projects in diverse areas such as low-code ML (Ludwig), distributed training (Horovod), probabilistic programming (Pyro), debugging (Manifold). Just this week, Uber released a new version of Orbit, a very innovative time-series forecasting framework based on Bayesian methods.

Uber’s contribution to the open-source ML space should not come as a surprise. After all, Uber has been running one of the largest ML infrastructures in the world, powered by their famous Michelangelo architecture. The importance and speed of Uber’s open-source ML contributions are undoubtedly impressive, but they aren’t an exception by any stretch. In the last few years, several tech firms like LinkedIn, Netflix, Airbnb, Lyft, and others have become highly active, open-sourcing several of the ML technologies they have incubated internally. Many can make the case that some of these open-source initiatives haven’t received the regular contributions and maintenance needed for mainstream adoption. However, it is unquestionable that those open-source releases have helped accelerate the innovation in large-scale ML architectures and pushed many ML startups to build on the foundation sets by these tech giants.

🔺🔻 TheSequence Scope is our Sunday free digest. To receive high-quality educational content about the most relevant concepts, research papers, and developments in the ML world every Tuesday and Thursday, please subscribe to TheSequence Edge 🔺🔻

🗓 Next week in TheSequence Edge:

Edge#157: we explore CI/CD in ML Solutions; we discuss Amazon’s continual learning architecture that manages the ML models lifecycle; we overview CML, an open-source library for enabling CI/CD in ML pipelines.

Edge#158: we finalize our MLOps series with deep dive into Aporia, an ML Observability platform.

Now, let’s review the most important developments in the AI industry this week

🔎 ML Research

Improving Reinforcement Learning with Lookahead Policy

Carnegie Mellon University published a paper detailing a technique to improve reinforcement learning agents with policies that look into the future to formulate better actions →read more on Carnegie Mellon University blog

Scaling Vision Transformers

Google Research published a paper detailing a mixture of experts (MoE) technique to scale the training of large vision models →read more on Google Research blog

Computer Vision for Amazon Product Pages

Amazon Research published a paper detailing a computer vision method used to identify and correct mistakes in its product catalog pages →read more on Amazon Research blog

🤖 Cool AI Tech Releases

Uber Orbit 1.1

Uber released the new version of Orbit, an open-source Bayesian time-series forecasting library →read more on Uber Engineering blog

🛠 Real World ML

Airbnb Conversational Agents

Airbnb published a blog post with insights about the architecture powering its conversational AI engine →read more on Airbnb blog

Data Science Experimentation at Netflix

Netflix published a new blog post providing more details about the architecture and techniques used to streamline experimentation across its data science pipelines →read more on Netflix Tech blog

Low Code ML at Ulta Beauty

Beauty products company Ulta Beauty details its approach to low code AI to improve the personalization of the user experience →read more in this coverage from VentureBeat

🐦 Follow us on Twitter, where we share all our recommendations in bite-sized form

💸 Money in AI

AIOps

Event correlation and automation platform BigPanda raised $190 million led by Advent International. Hiring remote, in the UK and US.
Unified observability and optimization platform Virtana raised $73 million in venture capital. Hiring in Poland and the US.

AI/ML/data

Data collaboration company Observable raised a $35.6 million Series B funding round led by Menlo Ventures. Hiring in San Francisco/US.
AI-generated synthetic data company MOSTLY AI raised a $25 million Series B round of funding led by Molten Ventures. Hiring in Vienna/Austria or remote.
AI automation company StageZero raised $1.8 million in a funding round led by Konvoy Ventures.

AI-powered

Sales enablement platform Highspot raised $248 million as part of a Series F round B Capital Group and D1 Capital Partners. Hiring globally.
Recruitment platform SeekOut raised $115 million in series C financing led by Tiger Global Management. Hiring in Bellevue/US or remote.
Location intelligence Placer.ai raised a $100 million Series C round led by Josh Buckley. Hiring in the US, Israel, and remote.
Autonomous building platform PassiveLogic raised $34 million in a Series B round led by Addition. Hiring in Holladay/US.
Medical transcription platform DeepScribe raised $30 million in Series A funding led by Nina Achadjian at Index Ventures. Hiring in San Francisco/US or remote.
Communication compliance platform Shield raised a $15 million Series A round co-led by Macquarie Capital and OurCrowd. Hiring remote.
E-commerce analytics and data company Daasity raised $15 million in Series A funding led by existing lead investor VMG Catalyst. Hiring in San Diego/US or remote.
Sales coaching solution Second Nature raised $12.5 million in a Series A investment from Signals Venture Capital, Stage One Ventures, Cardumen Capital, and Zoom’s Zoom Apps Fund. Hiring in Tel Aviv/Israel.

TheSequence