Machine Learning

Do Androids Dream of Electric Sheep?

Vision Language Action

version 2.9

Vision-Language-Action explores how modern VLA models can drive real robotic behavior by unifying perception, instruction, and control. This experiment evaluates SmolVLA, the LeRobot framework, and a pair of SO-101 3D-printed robotic arms, while comparing performance across heterogeneous compute platforms. The goal is to understand how multimodal inference workloads map onto different parts of an SoC, how execution characteristics vary between Intel and NVIDIA hardware, and what this means for deploying VLAs in real robotic systems.

Release Date: August 3, 2026

Core Concepts

Natural language + vision models controlling real robotic arms for precise manipulation.

Tic-Tac-Toe

version 1.5

Tic-Tac-Toe builds on the Vision Language Action work by introducing multiple interacting models that collaborate to play physical tic-tac-toe using the SO-101 robotic arms. This project explores how perception, reasoning, and control can be distributed across specialized components: one model interpreting the board state, another selecting optimal moves, and a third translating decisions into precise manipulation actions. By combining these capabilities, the system demonstrates how modular intelligence can outperform a single monolithic policy in tasks requiring both strategy and embodiment.

Release Date: August 18, 2026

Core Concepts

Multiple models working together to perceive, plan, and play tic-tac-toe physically.

Checkers

version 1.1

Checkers extends my multimodel research by orchestrating two SO-101 robotic arms as opposing players in a full physical game environment. Each arm operates as an independent agent, but both are driven by coordinated models that handle board perception, move selection, and piece manipulation. The project explores how strategic reasoning, turn-based planning, and embodied control can be unified to manage a complex, multi-step task with strict rules and spatial constraints. This system demonstrates how learned and deterministic components can work together to execute a complete game from opening moves to endgame.

Release Date: August 27, 2026

Core Concepts

Two robotic arms acting as CPU-driven agents to play a full game of checkers.

Precision Engineering

High-reliability embedded systems, FPGA design, and robotics platforms built with uncompromising attention to detail.

Verified Performance

Benchmarked, stress‑tested, and validated across real‑world robotics and compute workloads.

Robotics & Control

Physical systems, precise motion, and intelligent actuation.

Follow me on the socials!