WindBorne Systems is supercharging weather models with a unique proprietary data source: a global constellation of next-generation smart weather balloons targeting the most critical atmospheric data. We design, manufacture, and operate our own balloons, using the data they collect to generate otherwise unattainable weather intelligence.

Our mission is to eliminate weather uncertainty, and in the process help humanity adapt to climate change, be that predicting hurricanes or speeding the adoption of renewables. We are building a future in which the planet is instrumented by thousands of our microballoons, eliminating gaps in our understanding of the planet and giving people and businesses the information they need to make critical decisions. The founding team of Stanford engineers was named Forbes 2019 30 under 30 and is backed by top investors including Khosla Ventures.

WindBorne is looking for multidisciplinary Engineers with solid foundations in in Linear Algebra, Information Theory, and Computing to work on a variety of applied math and software projects to support our mission of building a Planetary Nervous System.

Here are some examples of projects you might work on:

1. Telemetry compression. Balloon telemetry is sent to the ground in 50-byte packets that each cost $0.06. That's over $1k per MiB, but thankfully each multi-week flight transmits far less than that with some decent custom compression that we use to save bits. But there is a lot of room for improvement, either thought a new hand-designed mathematical model to represent distributions, or possibly with a small Neural Network that is trained on prior flight data and then ran locally on a small ARM processor.
2. Balloon Guidance, Navigation, and Control. We simulate and optimize balloons using custom GPU kernels for integrating velocity from a wind field (from WeatherMesh predictions). We can simulate over 100,000 7-day flights in under 1ms. We use SGD to optimized flight profiles, and then uplink plans to balloons, and we are capable of getting balloons to within single-digit kilometers of targets routinely around the globe. There are many sub projects to work on here.
3. Balloon-to-Balloon mesh networking optimization. It's possible that in the future we will network our balloons to one another with balloon-to-balloon radio. This creates quite the fun and challenging network optimizaiotn problem: the connectivity graph is always changing in someone predictable and also controllable ways.

Here are some qualities for the type of person that we need. The below is written in first-person by me, John Dean, Cofounder and CEO of WindBorne

4. You mush be able to learn and understand a problem really fast, self-driven by curiosity. I can say to you "go fix the compression that we use on balloon telemetry" and then in two days you read through the code base with help of an LLM and have a few 1 on 1 conversations with other engineers, and you will be fully up to speed on the problem with a deep understanding. In that process, you likely asked over 100 questions to gain a very deep understanding very fast.
5. Every day you think about the following three things in some way: Linear Algebra, Information Theory, and Computers. You should think the term "neural network" is silly because it's just linear algebra with nonlinear activation functions thrown in the mix. You often think "how many bits of information do I need to represent that" when working on a problem, and the concept of "bit of information" is second nature to you. You should understand how fast computers actually are, and you are reasonably good at writing code that can really leverage compute. 5 milliseconds is quite a long time for a GPU (they render frames for 140Hz videogames).
6. Following from the above, you must be able to earnestly say "I love math" and "I love computers". I really love those two things and am very passionate about them. You also must truly love learning.
7. You must be able to play around and iterate fast through potential solutions to a problem. When you work on a new mathy/computing problem, your brain should expect dopamine of some kind of satisfying plot or result within the first day you look at it. You borderline might have ADHD, because you jump around between quick cool results a lot, but you have enough follow through to get things done
8. You must be able to work with computer systems and write production software. You are not someone who spends their time in Matlab, because it's not a real programming language for people who understand computers and software development. Also, it's probably a bad sign if you are overly attached to one language: LLM's make using new languages very easy and in the end, this work is not about the code, it's about how simply and efficiently orchestrate the information factory that is a computer.
9. You must know how to leverage modern LLM's to vastly increase the pace of software development, while also avoiding too much LLM slop in critical code paths.
10. You must be highly self-directed and self-motivated. You will start off reporting directly to me (John Dean). I have the background for this role, and I set up many of these mathematical foundations for things in WindBorne, and I spend 50% of my time getting involved in projects like this. I want to scale this skillset within WindBorne, but I'll be honest: I'm not sure that I will find the right person ever, or that this role that I have in mind even makes sense. We'll have to see!

‍

• 401(k)
• Dental insurance
• Health insurance
• Vision insurance
• Unlimited PTO
• Stock Option Plan
• Office food and beverages

• Palo Alto, CA, In person required.

‍