Arsam Aryandoust

(he/him)

I am a Postdoctoral Researcher at the MIT Laboratory for Information and Decision Systems, where my focus lies at the intersection of Artificial Intelligence (AI) and Climate Change. Before joining MIT, I completed my doctoral studies on AI for the renewable energy transition at ETH Zurich, a topic that continues to inform my current research endeavors.

Addressing Climate Change isn't just a scientific challenge but a profound moral obligation. It's about preserving billions of lives, safeguarding our ecosystems, and ensuring the continuity and flourishing of human society in its rich complexity. I firmly believe that AI must be harnessed if we want to solve Climate Change rapidly. Nevertheless, the number of experts currently engaged in this vital work compared with other application domains is alarmingly low. I am hence passionate about encouraging more engineers to join this global effort and provide them with research opportunities wherever I can.

My approach is to develop safe and secure AI software that can have an immediate impact in our battle against Climate Change. I believe that these solutions require a holistic and interdisciplinary approach, which makes me excited to be located at the new MIT Schwarzman College of Computing.

Scholar  /  Github  /  Python  /  Docker  /  Harvard Dataverse  /  Hugging Face 🤗  /  Bio

profile photo

Research

My research interest covers the broad field of developing and enhancing AI for tackling Climate Change. My focus, however, is currently on renewable power systems due to their urgency for mitigating greenhouse gas emissions. Recently, I have also started to learn more about the applications of AI in sustainable material discovery and negative emission technologies.

From Bitter to Better Lessons in AI: Embracing Human Expertise as Data
Arsam Aryandoust, Paul Liang
Coming soon, 2025

We argue that it is time to treat human expertise, whether written instructions, rules of thumb, equations, or code, as real data in the age of AI. When models leverage this know-how, such as physics principles, symmetries, or safety constraints, they often require far less training and generalize more effectively. Today’s large language models can already tap into such expertise through examples, feedback, or rapid retrieval. Yet, our survey of 1,000 NeurIPS papers reveals that while LLM use is growing, few works deeply integrate expert knowledge, leaving a major opportunity for building AI that is more efficient, reliable, and cost-effective.

Artificial Intelligence for the renewable energy transition
Arsam Aryandoust
ETH Zurich, 2023

Artificial Intelligence and its subfield of Machine Learning are excellent tools for expressing the world in numbers, discovering their patterns, and designing solutions based on these in higher dimensions than visible to our human eyes and for higher complexity than understandable for us humans otherwise. We show that this also holds for transforming energy and tackling climate change.

Enhanced spatio-temporal electric load forecasts using less data with active deep learning
Arsam Aryandoust, Anthony Patt, Stefan Pfenniner
Nature Machine Intelligence 4, 977-991, 2022
Github / PyPI / Docker / Dataverse

We develop an active deep learning algorithm that makes better spatio-temporal predictions of electric load with less data compared to traditional passive deep learning algorithms.

City-scale car traffic and parking density maps from Uber Movement travel time data
Arsam Aryandoust, Oscar van Vliet, Anthony Patt
Scientific Data 6, 158, 2019
Github / PyPI / Docker / Dataverse

We develop a Hidden Markov Model that is able to infer accurate city-scale car traffic and parking density maps from Origin-Destination travel changes.

The potential and usefulness of demand response to provide electricity system services
Arsam Aryandoust, Johan Lilliestam
Applied Energy, 204 (15), 749-766, 2017

We demonstrate why Demand Response technology has remarkable potentials for providing electricity system services on short time-scales, but not on long time-scales.

Athletics

I enjoy doing indoor and outdoor sports to balance out long days of work at the computer. I also enjoy exploring new disciplines on a regular basis.

Boxing

The sports that I am most passionate about is boxing. I started boxing at the early age of 13, which is the reason I consider it to be my main athletic discipline. Recently, I have also started to learn how to throw and defend kicks, and become more familiar with elements of kickboxing.

Calisthenics

Derived from the Greek words "kallos" (beauty) and "sthenos" (strength), calisthenics embodies the elegance of physical strength. What excites me most is its seamless blend of simplicity and sophistication. With just your own body weight, you can explore an endless variety of exercises that transform each workout into an opportunity for growth and creativity. Beyond the physical challenge, calisthenics allows me to connect with a passionate community of like-minded folks wherever in the world I exercise. People, who share a passion for learning, exchanging stories, and inspiring one another both in exercise and beyond.

Rowing

I started learning rowing after moving to Cambridge, where the sport is deeply woven into the local culture. What began as a curiosity has slowly grown into a passion. I especially enjoy it in the summer, when early mornings on the water provide a quiet yet demanding start to the day. Being out on the river at 5 am sharp brings a discipline that pulls me out of bed and sets the rhythm for everything that follows. Rowing challenges both coordination and endurance, but what I value most is how it structures the day with a sense of calm and purpose.

Running

Running, for me, is the most fundamental form of movement. It requires no equipment, only the willingness to set one foot in front of the other and keep going. I enjoy the freedom it brings, being able to run anywhere, anytime, turning cities, trails, or waterfronts into my training grounds. Running, for me, is also a way of building discipline: pacing myself, pushing through fatigue, and finding joy in incremental progress. Just as importantly, it serves as a way to clear my mind and reflect, often leading to fresh ideas and perspectives.


Credits for the design of this website go to Jon Barron for making the source code of his website publicly available.