Memory Leak
Published in

Memory Leak

Memory Leak — #6

VC Astasia Myers’ perspectives on machine learning, cloud infrastructure, developer tools, open source, and security. Sign up here.

🚀 Products

Devbox

Devbox is a command-line tool that lets you easily create isolated shells and containers. You start by defining the list of packages required by your development environment, and devbox uses that definition to create an isolated environment just for your application. In practice, Devbox works similar to a package manager like yarn – except the packages it manages are at the operating-system level (the sort of thing you would normally install with brew or apt-get).

Why does this matter? Devbox offers some of the benefits of Nix, a well-known tool for package management and system configuration. Nix is known for being complex so Devbox helps make it more accessible. Devbox is internally powered by Nix.

Cal.com

Cal.com is an open source alternative to Calendly. This week Cal.com released v2.0 that included a rebuilt of the UI from scratch. It is now free for individuals.

Why does this matter? Cal.com plays into the theme of open source at the application layer. There is an enduring question of how effectively these types of startups can monetize their users. Often application layer open source businesses have struggled because designers aren’t part of the community and don’t contribute to the solution’s advancement. In turn, it is particularly interesting that Cal.com did a redesign.

📰 Content

MLOps at Walmart

Walmart discusses some insightful details about the MLOps infrastructure and processes powering their ML pipelines.

Why does this matter? One of the key challenges Walmart highlighted was data drift, which is when the statistical properties of the target variable, which the model is trying to predict, change over time in unforeseen ways. This causes problems because the predictions become less accurate as time passes. Cata-centric debugging tools like Unbox and monitoring like Arize help identify data drift.

Lessons Learned: The Journey to Real-Time Machine Learning at Instacart

Instacart highlights the difference between batch-oriented ML systems and real-time ML systems. The paper focuses on features and serving. Most machine learning in production is about leveraging signals (features) derived from raw data to predict targeted goals (labels). Batch features are more stale than real-time features. Moving to real-time serving means serving precomputed predictions to real-time serving in order to reduce staleness, limited coverage, and resource underutilization. Transitioning to a real-time serving system has been made possible by two products: Feature Store and Online Inference Platform.

Why does this matter? Instacart gave an example a typical shopping journey, powered by hundreds of ML models. All of these actions happen in real time, which means leveraging machine learning in real-time can provide significant value to the business. We are seeing an increasing number of ML teams move to real-time ML infrastructure to better suite real-time use cases. The transition from batch to real-time has challenges so we are starting to see businesses like Claypot.ai arise to solve the pain points.

Instacart’s ML Platform with Real-Time Serving and Real-Time Features

The 5 Phases of Figma’s Community-Led Growth: From Stealth to Enterprise

In this interview, Claire Butler, a Senior Director of Marketing at Figma, imparts lessons about how to build and cultivate a community along each phase of the startup journey — from the earliest innings of the company, all the way to bringing on a sales team and targeting more enterprise deals. She shares the specific creative tactics Figma used to energize the design community and build organic momentum when the product was just beginning to take shape. Butler flags some of the key decisions that paid off along Figma’s journey — including making the call to finally emerge from stealth, introducing pricing with the right gating strategy and finally bringing in a sales motion.

Why does this matter? Figma has built an incredible product and community. This week Adobe announced that it would acquire Figma for $20 billion, taking out one of its biggest rivals in the realm of digital design. Figma acquired at 50x annualized forward revenues and is one of the largest software acquisitions ever. Clearly they did something right and this piece highlights some useful tactics to learn from.

💼 Jobs

⭐️Claypot Founding engineer (infra)

⭐️HumanitecBackend Software Engineer (fully remote)

⭐️DiagridFrontend Engineer (remote)

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Astasia Myers

Founding Enterprise Partner @ Quiet Capital, previously Investor @ Redpoint Ventures and Cisco Investments