External 
AWS Public Sector
AWS Public Sector EPFL designs robots through artificial evolution
AWS Public Sector feature on the Robogen platform. Co-evolving robot bodies and brains via genetic algorithms, parallelised on AWS.
Robogen: robot bodies and brains co-evolving on AWS
Open-source evolutionary robotics platform, re-architected to scale simulation across EC2.
C++ Python AWS EC2 autoscaling Genetic algorithms Neural networks Webots
Problem
Evolutionary robotics needs massive parallel simulation: thousands of generations, dozens of robot designs, each rebuilt in physics from scratch. The lab pipeline ran on local hardware that evaluated one population at a time. Class projects queued behind research runs. The infrastructure throttled both the science and the teaching.
What I built
I split Robogen into two cooperating services: an evolution engine (population, selection, mutation, reproduction) and a physics simulator for fitness evaluation, talking over the network so the simulator could scale on its own. I moved the simulator to AWS with autoscaling, peaking at 15 EC2 instances during class demand, and secured AWS research credits to pay for the compute. As TA on the EPFL Evolutionary Robotics course I taught the platform and supervised 20+ master's students through end-to-end experiments. The best designs printed on the lab's 3D printer and ran on off-the-shelf electronics.
What I learned
Separate the search from the evaluation. Once the expensive part scaled independently, 100+ students could run real experiments. The search and embodied-learning intuitions from this work still shape how I design agent systems.