PROLOGUE — SYSTEM ONLINE
> ▌
I'm Ashok GC — Robotics & Mechatronics MSc engineer at the University of Leeds, with 3+ years across the full stack: control design, kinematics, sensor fusion, machine learning and embedded development from bare-metal C on ARM to Verilog RTL on FPGA. Scroll to live the journey — and at the end, play it.
CHAPTER 01 — KATHMANDU, 2020–2023
The spark wasn't a single moment — it was a question that wouldn't let go: could code make something in the real world move on its own?
I chased it through a BSc in Computing at Softwarica College (Coventry University, UK), graduating First Class Honours. The college Innovation Hub became my workshop — a Mars-rover replica, a drowsiness-detection system, anything that let software meet hardware.
It all converged in my thesis: a connected autonomous vehicle with CNN-based object detection, lane tracking and V2V communication — ~95% detection in testing. It was awarded Best Thesis 2023. The boundary between code and matter had dissolved for good.
"I didn't want to just write programs. I wanted to write things that could see, decide and move."
CHAPTER 02 — APPLIED AI, 2023–2025
A degree teaches you what works in theory. Industry teaches you what survives contact with real data, real deadlines and real users.
As an R&D Software Engineer at URP & Associates, I designed and deployed AI/ML solutions that automated audit and document-processing workflows — cutting manual review effort and taking applied-AI initiatives from concept all the way through to deployment.
Earlier, as a Research Assistant at the Softwarica Innovation Hub, I built ML models across AI, robotics and IoT — including a humanoid robotic arm — and mentored student project teams. Two years of making models actually run in production sharpened everything that came next.
# from prototype to production
model = load("doc_classifier.pt") # trained on real audit docs
pipeline = Pipeline([ingest, ocr, model, route])
for batch in stream(audit_queue):
result = pipeline(batch)
review_time = result.saved_hours # manual effort, gone
✓ Deployed. Monitored. Iterated.
✓ Concept → production, end to end.CHAPTER 03 — THE FORGE, LEEDS 2025–2026
An MSc (Eng) in Robotics & Mechatronics at the University of Leeds — where the self-taught instincts finally got the equations to match.
Control Systems Design, Aerial Robotics, Biomechatronics & Medical Robotics, FPGA Design for SoC, Embedded Microprocessor Systems — a year spent going deep across the whole stack, from Riccati equations to Verilog RTL.
My individual project is the centrepiece: a two-wheeled self-balancing and jumping robot — a gain-scheduled LQR balance controller and a full jump state machine, simulated in MuJoCo at 500 Hz. That's the thread through all of it: make it balance, then make it leap.
LQR Riccati solved at multiple leg heights.
FSM Crouch → power stroke → flight → land.
SIM MuJoCo loop locked at 500 Hz.
GOAL Balance, then jump the obstacle. 🤖
CHAPTER 04 — THE BUILD LOG
From control theory and CAD to perception and bare-metal firmware — a year of building across the whole robotics stack.
MSc individual project: a two-wheeled balancing robot with an actuated jumping mechanism. Gain-scheduled LQR balance controller (Riccati solved across leg heights) and a crouch→power-stroke→flight→land state machine, simulated in MuJoCo at 500 Hz.
Hexacopter with a 5-DOF manipulator and parallel gripper to autonomously retrieve floating debris from urban canals. Sized for thrust-to-weight > 2 and ~12-min endurance; Jetson Nano + YOLOv8 perception with image-based visual servoing. Lead author of the IEEE-format paper.
Preliminary design of a 3-DOF ankle-rehab exoskeleton (plantarflexion/dorsiflexion, inversion/eversion, ab/adduction). SolidWorks model with FEA validation, forward/inverse kinematics and Jacobian analysis, plus a hierarchical patient-cooperative control architecture.
Conceptual design of a three-arm, 7-DOF telesurgical system for remote keyhole surgery. Bilateral master–slave control — adaptive fuzzy-tuned PID, passivity & impedance control with model-predictive compensation for ≤200 ms latency — plus tactile slip detection and haptic feedback.
ANN, SVM and CNN classifiers in MATLAB recognising five gait activities from wearable IMU data for an intelligent lower-limb prosthesis. Feature extraction with Bayesian hyperparameter tuning — CNN 97.9%, SVM 93.8%, ANN 89.7% test accuracy.
Real-time interactive Mandelbrot/Julia explorer on a DE1-SoC, driving an LT24 LCD at ~19 fps from a 50 MHz clock. Custom parameterised Q4.20 fixed-point RTL, 12 zoom levels (~4096×), verified in ModelSim and validated on hardware.
CHAPTER 05 — THE ARSENAL
Hands-on across the full robotics stack, distilled into four core capabilities.
Making machines balance, move and decide — from the maths to the motors.
Perception and decision-making — models that see, classify and generalise.
Down to the bare metal — firmware, RTL and the boards they run on.
The languages and toolchain that tie the whole system together.
CHAPTER 06 — PRESENT DAY
School of Electronic & Electrical Engineering (expected Sept 2026). Control Systems Design, Aerial Robotics, Biomechatronics, FPGA & embedded systems. Individual project: a two-wheeled self-balancing & jumping robot.
Designed and deployed AI/ML solutions automating audit and document-processing workflows, cutting manual review effort. Built reliable data pipelines and internal tooling — concept through to deployment.
Interdisciplinary research across AI, robotics and IoT — including ML models for a humanoid robotic arm. Mentored and supervised student project teams.
Thesis: “AI and ML for Autonomous Vehicle Development” — a connected autonomous vehicle with object detection, lane tracking and V2V communication. Awarded Best Thesis 2023. Where it all began.
FINAL CHAPTER — PLAYABLE
The story doesn't end with me. ARC-7 — the little bot you've watched assemble itself this whole way down — has gained sentience and is making a break for freedom. Crates block the corridor, security drones patrol overhead. Help it escape. How far can you run?
// Yes, I coded this too. The physics engine is 9 lines. The fun is free.
EPILOGUE — HANDSHAKE PROTOCOL
I'm seeking a graduate robotics or mechatronics role from late 2026. Got a robot to build, a controller to tune, or a perception problem to crack? My inbox is always open.