Neuroscience to AI to climate: Archy de Berker, Climate TRACE

Archy de Berker is no stranger to career transitions. Trained as a neuroscientist, he became a data scientist after attending a data science bootcamp. After a few years at an AI startup, Archy began looking for more meaningful work.

He did contract work with a number of renewable energy companies (e.g., Origami Energy and Habitat Energy), and just recently began a contract with Climate TRACE, choosing it over a lucrative full-time offer from another AI startup.

Q: What made you decide to look for work on climate?

It came down to taking a close look at what I thought and said was important, and comparing it to the actions I was taking. If people like me with science PhDs don’t make their life choices according to evidence, then who will?

Then my dad asked me: how would a 12-year-old Archy choose between automating tech support and working on the greatest problem of our time? For him, it would be a no-brainer. And once I realized how abundant climate roles are, it was a no-brainer for me too.

Finally, in light of BLM events, I thought a lot about my privilege. A positive thing to do with it is to work on problems and take risks that others cannot afford to take.

Q: Were there any sacrifices involved in making this choice?

Engineers and data scientists in climate are paid well. But there is a bit of a “sexiness” sacrifice: climate startups are not yet seen as sexy as AI startups — though some companies, like Pachama, do a good job of being both.

The other sacrifice is becoming a beginner again. So far my climate work has been about 5% ML and 95% domain expertise, which I’m learning on the job: I spent the last week educating myself on carbon emissions of heavy industry, and today I wrote a document on how paper is manufactured.

Q: How can one get a handle on this domain coming from a different background?

I want to acknowledge that this is pretty overwhelming and difficult. Some resources I found really useful are A Pragmatic guide to Climate Change from Tomorrow and Ryan Orbuch’s blog post about carbon removal. I’ve also found podcasts to be helpful — I love Gimlet’s How to Save A Planet, which is very accessible.

Q: What strategies did you use to find the right climate work for you?

I just talked to a lot of people. Eventually I got introduced to a climate VC in Montreal. Such people know a disproportionate amount of other people and opportunities, and before long, he shared on LinkedIn a job with Climate TRACE, which turned out to be a great fit for me.

Q: What is a specific climate problem you worked on?

Right now I’m coming up with a plan for ClimateTRACE to monitor emissions from heavy industry. For many countries, hitting emissions targets critically depends on this sector.

An ML model trained on data from countries with high-quality monitoring standards can generalize to countries with poor-quality data, allowing us to monitor emissions at the satellite revisit rate (5 days), which is much better than annual reporting.

Satellite data is useful for this, and it’s interesting figuring out how to apply classic ML models to the problem. For example, some factories have parts that get really hot and you can spot them from a satellite to know when the factory is running. Paper factories have massive log piles next to them — can we estimate their size to predict factory throughput?

Q: What misconceptions do people in tech have about working on climate?

One is “There are no good jobs in climate.” I’m not sure if this was ever true at all, but it’s definitely becoming less and less true over time.

Another is “The problems to solve are not software problems; if only I had been a battery engineer instead.” A lot of the hard science problems (such as solar efficiency) have been largely solved — so today almost every organization working on climate needs good software engineers and data scientists. We need to come in and do what we do best — make the hard science easy to use at scale!

Q: What’s next for you?

I’ll probably be moving back to the UK to find for-profit work in climate tech in London, where most of the UK climate startups are. I think we’re at a point where the right policies and incentives are in place to allow for-profit companies to take risks and attempt to scale fast, which is what the startup ecosystem is really good at. This episode of How to Save A Planet gives a wonderful overview of how this can work.

I’m excited about carbon marketplaces — such as the work that Pachama and Indigo Ag are doing — I really like that these platforms connect people who have the means to sequester carbon with companies who have the funds to do it.