Building autonomous systems from first principles, one experiment at a time.
Notes on robotics, AI, autonomy, and drones — documenting how machines perceive, decide, and move through the world. A personal archive of prototypes, failures, and the physics behind the code.
Engineering is a research practice, not a deliverable factory..
This site is a working notebook — hypotheses tested in hardware, models validated in simulation, and systems shipped only after the underlying physics makes sense. Robotics, AI, and autonomy treated as one continuous inquiry.
Read the manifesto
to control loops, I
document the full
stack of autonomous
systems.
Four domains, one thread of first-principles inquiry.
Each domain is a lens on the same question: how do you build machines that understand their environment and act reliably in it?
Robotics
& Manipulation
Kinematics, dynamics, and embodied intelligence — from arm trajectories to mobile platforms in unstructured environments.
Autonomous
Systems
Planning, decision-making, and closed-loop control for agents that operate without constant human supervision.
A living archive of experiments in motion, perception, and control.
Quadrotor Dynamics
First-principles model of thrust, torque, and attitude control — from equations to hover in simulation.
Visual Odometry
Feature tracking and pose estimation on constrained hardware — bridging classical CV and learned descriptors.
Path Planning
Graph search and trajectory optimization for cluttered environments — safety margins as a design constraint.
From observation to working systems.
Every stage is iterative, measurable, and grounded in physics — no black boxes, no hand-waving.
Observe →
Define the problem in physical terms. What are the constraints, sensors, actuators, and failure modes?
Model →
Build a mathematical or simulation model. Validate against real data before writing a single line of production code.
Build →
Implement incrementally — smallest testable unit first. Log everything; reproducibility is non-negotiable.
Validate
Test in simulation, then hardware. Compare predicted vs. measured. Iterate until the gap closes.
Autonomous Drone Stack
End-to-end aerial autonomy — perception, planning, and control integrated on a custom quadrotor platform with full telemetry logging.
Robotics Simulation Lab
A reproducible simulation environment for testing control algorithms before hardware deployment — Gazebo, ROS, and custom physics models.
“The best engineering starts with understanding the physics — not with picking a framework or chasing the latest model release.”
Influenced by the engineers and researchers who publish their failures alongside their wins.
Read more notes
Let's build something autonomous and grounded in first principles.
Open to collaborations on robotics, drones, and autonomy. Reach out via email or GitHub — always interested in rigorous engineering conversations.
