Bakel Bakel

I build robots that work in the real world not just in controlled lab conditions, I can work in labs and fields, under pressure, in front of live audiences, at 500 Hz, with milliseconds to spare.

My work sits at the intersection of three things most engineers only touch one of: AI (reinforcement learning), embedded systems, and industrial robotics infrastructure. I don't specialise in the gaps between them. I own all of it.

I've trained RL policies that taught a Unitree Go2 Edu 43 degree-of-freedom humanoid to walk, balance dynamically, and kick with precision, then implemented the full sim-to-real pipeline myself, from MuJoCo system identification and actuator torque curve validation to domain randomisation and hardware deployment, landing under 8% performance degradation from simulation to physical hardware. I've built the ROS 2 production stack it runs on, written the BehaviorTree orchestration across seven concurrent behaviours, deployed the safety supervisor that kept a live crowd safe, and integrated the YOLOv8 perception pipeline that let the robot see and react in real time.

At BiaTech, I architected BiaLogic, a multi-robot mission control platform coordinating six heterogeneous robots simultaneously across industrial inspection sites and delivered 99.7% uptime. I also built the full-stack React.js dashboard operators used to monitor all of it live. I wrote the bootloader configs and resolved the firmware conflicts that made the Jetson platforms boot reliably in the first place. That's the range I operate at.

I've tuned PID and LQR controllers for underwater ROVs in tidal currents. I've diagnosed MAVLink firmware failures at 2am and fixed them. I compiled OpenCV from source on ARM Linux with a GStreamer backend because the prebuilt binaries weren't good enough.

I collected 4,000 human demonstration trajectories through a glove I instrumented myself to seed imitation learning pipelines. I won an international autonomous boat race. I designed and shipped a 4-layer robotics PCB.

I resolved a BLE firmware bug in under 48 hours that had blocked another team for weeks.

In parallel with all of this, I founded Aurora Robotics, now the most active robotics organisation on the African continent, growing to 32 staff, 11 live research projects, and 1,200+ students across 30 Nigerian universities within six months, with no external funding.

I hold a First Class Bachelor's in Marine Engineering, Best Engineering Student four consecutive years and a Double Master's in Marine Engineering with Specialization in Autonomous Underwater Robotics on a fully-funded Erasmus Mundus Scholarship, 106/110 GPA, across universities in Spain and Italy.

I'm Top Rated on Upwork with a perfect 5.0 across robotics, ML, and embedded systems.

I build things that ship. Then I go back to the hardware and build the next one.

At BiaTech Corporation in the United States, I was the sole architect and lead engineer for BiaLogic, a modular ROS 2 mission control platform supporting heterogeneous robot fleets AgileX, Unitree G1, UFactory, Yahboom, and Skydio robots, across industrial energy-sector inspection deployments. I deployed complete ROS 2 software stacks across four-plus NVIDIA Jetson edge platforms (Orin Nano, Xavier NX, and Nano), achieving stable real-time policy execution at under 50 milliseconds sensor-to-actuator latency across six or more simultaneous robots with 99.7% uptime across all multi-robot industrial deployments. I reduced system initialisation time by 35% through boot sequence optimisation and service dependency tuning on Jetson platforms.

At CITIC, the research centre of the Universidade da Coruña, and in partnership with Ardentia, I engineered a production-ready AI computer vision pipeline for autonomous underwater pipeline detection using ROV imagery, achieving greater than 85% detection accuracy in near-zero-visibility marine conditions. I led data collection, annotation, and model training for a 2,000-plus frame polygon and mask segmentation dataset, fine-tuning YOLOv8 to 99.5% mAP50 and 99.1% precision at approximately 3 milliseconds per frame on GPU inference. I developed, tuned, and validated PID and LQR controllers for depth hold and 6-DOF stabilisation of the BlueROV2 in strong tidal currents, achieving ±5 cm steady-state depth accuracy at currents up to 0.8 m/s.

I built a GStreamer-based real-time video pipeline on ARM Linux by compiling OpenCV from source with the GStreamer backend, streaming live underwater visual data at 30 fps to ROS 2 topics. I debugged and resolved MAVLink and DroneKit communication failures at the firmware level and authored a custom Python abstraction library over DroneKit, automating arming and disarming sequences and reducing BlueROV2 field-deployment time by approximately 50%. I also applied systematic system identification to characterise thruster dynamics and update simulation models.

I am Top Rated on Upwork with a 5.0 rating across all engagements.

Delivered contracts include: a full ROS 2 navigation stack with RTAB-Map and gmapping SLAM, Nav2 path planning, LiDAR and depth camera sensor fusion, and obstacle avoidance, rated 5.0 "Committed to Quality" ($1,554 earned); a complete ML pipeline for IMU signal classification covering preprocessing, feature engineering, model training, and embedded inference, rated 5.0 "Went above and beyond expectations... worked through challenges extremely well" ($1,300 earned); and a full ROS 2 swarm robotics framework with Gazebo and Isaac Sim simulation, PCB integration, and drone/AMR control, rated 5.0 "Unmatched dedication, consistency, and technical brilliance" ($1,300 earned).

Double Master's Degree in Sustainable Ships & Shipping 4.0 (Autonomous Underwater Robotics) (Erasmus Mundus) Universidade da Coruña, Spain & Università degli Studi di Napoli Federico II, Italy.

GPA: 106/110. 2024–2026.

Bachelor of Engineering in Marine Engineering (First Class) — Federal University of Petroleum Resources Effurun, Nigeria.

GPA: 4.77/5.0. 2016–2022.