Elias
Tragas.
Research-oriented software engineer building the future of physical AI — from recommendation systems to autonomous robots and embodied intelligence.
Engineering intelligence
into the physical world
Elias builds systems that help machines understand, learn from, and act in the real world. His work spans ML infrastructure, large-scale recommendations, NLP, and autonomous robotics — from Google-scale systems to early-stage startups pushing the boundaries of embodied AI.
Currently at Generalist AI, he's part of the team developing general-purpose foundation models for robots — enabling machines to learn from real-world data and adapt to new tasks with remarkable efficiency.
Building robots that
make headlines
Elias is part of the Generalist AI team whose work was recently featured as a Forbes Daily Cover story — profiling how the company is applying AI scaling principles to physical robotics.
Daily CoverAreas of expertise
Autonomous Robotics
Embodied foundation models, physical AI, autonomous vehicles, and real-world perception systems.
Machine Learning Infrastructure
Distributed training at scale, ML pipelines, model serving, and production systems.
NLP & Transformers
Transformer architectures, mixed-precision training, long-sequence models, and classification systems.
Recommendations
Large-scale recommendation engines, probabilistic search, and real-time inference in C++ and Python.
A decade of building
intelligent systems
University of Toronto
B.Sc. — Computer Science (A.I. Specialist) & Cognitive Science · Math Minor · 2013–2017
Undergraduate researcher and NSERC CREATE-UAV recipient. Built a constant-time inference recommendation platform using mutually recursive RNNs. Research on autonomous indoor mapping with MAVs and neural visual odometry under professors David Duvenaud and Raquel Urtasun.
Beyond ML, Elias has deep interests in cognitive science, philosophy of mind, and entropy-based models of life.
Teaching assistant for Machine Learning (CSC411), Mathematical Reasoning (CSC165), and Computational Thinking (CSC104).
Research
Things built for fun
& for the field
AutoDrone
Autonomous drone research platform built in Python for indoor mapping with MAVs.
GPSDrone
GPS-based autonomous drone navigation system.
Track Car #68
Races a 2015 Scion FR-S at Laguna Seca with SpeedSF. Also competed in the legendary 24 Hours of Lemons endurance race.
Burning Man
Regular on the playa — where art, community, and radical self-expression meet the desert.
EEG Light Show
A light show driven by brainwave signals via EEG.
SSH Synthesizer
A synthesizer controlled remotely through SSH.
Rajawali 3D Engine
Contributed to an Android OpenGL ES 2.0/3.0 3D rendering engine.
What matters
Let's connect.
Or send a message directly: