Physical AI · Computer Vision · Applied Mathematics
- 🎓 Engineer (M.Sc.) in Mathematics & Modelling, National School of Mathematical Engineering and Modelling, Benin
- 🛰️ I build AI that works outside the lab: embedded vision for autonomous drones, UAV crop monitoring, and optimization for real systems
- 🔬 Interested in Physical AI: bringing intelligence to systems that deal with sensors, motion, and imperfect real-world data
- 🗣️ Trainer & mentor: I like making complex models understandable to the people who depend on them
- 📫 Reach me at fructueuxhollo@gmail.com
- 🖼️ Computer Vision: segmentation, classification, attention mechanisms, transfer learning on RGB + infrared imagery
- 🛸 Embedded & Autonomous Systems: onboard vision for UAVs, coverage path planning, simulation-in-the-loop
- 🧮 Optimization & Operations Research: metaheuristics, constraint programming, scheduling
- 📊 Applied Statistics & Modelling: time series, geospatial analysis, surrogate models
Languages
Machine Learning & Vision
Optimization & Scientific Computing
Tools
🛸 Physical AI & Autonomous Systems
- Drone Coverage Path Planning: Coverage path-planning algorithms in Python, validated on a simulated drone (MATLAB UAV Toolbox): 100% coverage, 0 collisions, 2.625 Jerk RMS.
- Autobalancing Robot: Controller design and parameter optimization for a self-balancing inverted-pendulum robot, with a benchmarking suite and simulation visualizer.
🖼️ Computer Vision & Remote Sensing
- Remote Sensing in North Benin: Transfer learning for aerial object and activity recognition on a heterogeneous RGB + infrared dataset, 1400+ images self-collected with a DJI Mini 4 Pro, mAP@50 = 91.85.
🔊 Signal Processing & Dynamical Systems
- Sound ex Machina: Audio fingerprinting engine (Shazam-style): spectrogram peak constellations and hashing to identify a music track from a short query sample. Built in MATLAB.
- VortexCrypt: Image encryption scheme driven by a reversible Gray–Scott reaction–diffusion model, using the system's dynamics to diffuse and recover image data.
🧮 Optimization & Decision Systems
- Ant Colony Optimization for Task Assignment: ACO on a bipartite job/worker graph (27 jobs, 6 workers) for static scheduling in grid computing, reducing makespan by 53.6% vs Local Search and 32.9% vs a Genetic Algorithm.
- Timetabling Solver: Constraint-programming timetable scheduler built with Google OR-Tools.
- RL Maze Solver: Reinforcement learning environment and agent, with the environment implemented from scratch.
📊 Applied Machine Learning
- Insurance Damage Pricing: Non-life insurance pricing model combining a frequency model (MAE 0.27) and a severity model (MAE 0.09).
- 🎤 Oral presentation at ICCCS 2026 (11th International Conference on Computer and Communication Systems), Wuhan, China — UAV Data Acquisition and Multi-Stage Deep Learning for Maize Foliar Disease Detection: U-Net segmentation (>75% IoU) feeding a ResNet classifier (96.9% accuracy), cutting the image area to classify by 68%
- 🥇 1st place: TotalEnergies EEIA 2023 Machine Learning Challenge (thermal power output forecasting, 1st of 120+ participants)
- 🥈 2nd place: MTN MoMo Hackathon 2023, for SimpliMoney, a financial-inclusion app for illiterate users
- 🧑🏫 UNDP AI Bootcamp facilitator: mentored 54 innovators from 9 countries
- 🎓 Master's thesis defended with 89/100
