Sardor Israilov

Engineer and PhD in AI and robotics, based in Nice, France.

I teach robots to move with deep reinforcement learning, and I build the C++ runtimes that run industrial automation. My thesis trained a robotic fish to swim faster using camera and force-sensor feedback.

Download CV (PDF) See research Email me

Portrait of Sardor Israilov in a blue polo shirt

Now: R&D runtime engineer, UniversalAutomation.org
Thesis: Robotic fish control with deep RL, 2020–2024
Languages: English, French, Russian, Uzbek, German (A2)

About

I hold a PhD in AI and robotics from Université Côte d’Azur, done at the I3S, Inria and INPHYNI labs. It covered three systems: a cart-pole, a static robotic fish and a free-swimming robotic fish, each with its own physical model.

Before that I earned an MS in image and signal processing at INSA Lyon, with a three-year apprenticeship at Solystic (Northrop Grumman) in industrial automation and AGV motorization.

Today I work at UniversalAutomation.org, a community founded by Schneider Electric. I write the C++ runtime for IEC 61499 PLCs and maintain its build system. Before that I wrote Go microservices and Kubernetes deployments at Schneider Electric.

Watch the PhD defense (video) Read the thesis summary (PDF)

Research

Robotic fish swimming setup and learning curves

Learning to swim fast with deep reinforcement learning

Robotic fish · 2020–2024

A deep RL agent optimizes the swimming gait of a robotic fish from a camera, force sensors, an IMU and a Raspberry Pi, first in simulation and then on the real robot.

Visual servoing with a VAE coupled to deep RL.
Speed and thrust optimization of an autonomous fish.
Simulation with Gym, MuJoCo and COMSOL, plus real experiments.

Python · PyTorch · Gym · MuJoCo · COMSOL · OpenCV

Paper (PDF) Demo video

Robotic fish thrust measurements and model

Optimal control for maximum thrust in a robotic fish

Physical modeling · submitted to Nature Communications

Dynamics modeling, identification and simulation of the fish, then control strategies that maximize thrust. I built the hardware from CAD models and 3D prints.

System identification against force-sensor data.
CAD design and 3D printing of the fish.
Thrust optimization with fluid-dynamics simulation.

COMSOL · SolidWorks · MATLAB

Preprint (arXiv) Poster (PDF)

Cart-pole reinforcement learning results

Cart-pole control with deep RL, for teaching

I3S with Inria · PLOS ONE 2023

A general framework for learning control of an inverted pendulum, in simulation and on a physical setup. I also taught deep RL workshops in PyTorch with it.

Dynamics identification and sim2real transfer of the policy.
Mechanical design and installation of the experimental rig.

Python · PyTorch · SolidWorks

Article (PLOS ONE) Code Demo video

Medical image segmentation with Mask R-CNN

Master’s thesis · 2020

Automatic segmentation of heart scans, trained in the cloud.

Code and report

Publications

Submitted to Nature Communications. Optimum control strategies for maximum thrust production.
PLOS ONE. Reinforcement learning approach to control an inverted pendulum: a general framework for educational purposes.
IEEE Transactions on Ultrasonics. LU-Net: a multistage attention network to improve the robustness of segmentation.
In preparation. Learning to swim fast with deep reinforcement learning.
All papers on Google Scholar

Experience

Feb 2025 – now

R&D runtime engineer, IEC 61499 PLC

UniversalAutomation.org, created by Schneider Electric

Develop the C++ runtime, including new cybersecurity and packaging features.
Release runtime features for the UAO community.
Engineer the build system (Perl, Makefile, MSVC).
Run trainings and demos on deploying and configuring the runtime.

Apr 2024 – Feb 2025

R&D engineer

Schneider Electric, EcoStruxure™ Automation Expert

Wrote Go microservices for fleet management.
Deployed the containerized (Docker) PLC runtime with Kubernetes.
Refactored the existing codebase with new cybersecurity measures.

Oct 2020 – 2024

PhD researcher, robotics and deep RL

I3S / INPHYNI, Nice · DS4H selective scholarship

Deep learning for swimming optimization of a robotic fish, using a camera, force sensors, an IMU and a Raspberry Pi.
Dynamics modeling, identification and simulation of the fish.
Deep RL control of a cart-pole in simulation and on a real setup.

Sep 2017 – Sep 2020

Apprentice engineer: robotics R&D, then industrial automation

Solystic (Northrop Grumman), Alixan

Robotics R&D, AGV motorization (2019–2020)

Motorization study with real-robot tests, data analysis and simulation in Python and MATLAB/Simulink.
Empty-tray detection with Python and C++ OpenCV.
Preliminary CAD of a swerve drive in CATIA V5.

Mechatronics R&D (2017–2019)

PLC function blocks in CODESYS and SoMachine for automation and safety (Schneider M251).
CODESYS FBD library for roller-motor conveyors over AS-i and Modbus, with simulation.
MATLAB sizing tools for choosing electric motors and drives.
Installed postal-sorting machines in Bologna and Rome (Jun–Aug 2019), qualifying safety equipment with SoMachine and SICK Flexisoft.

Jul – Aug 2016

Technico-commercial intern

Vertiv (Emerson Network Power), Tashkent

Started up and maintained UPS units and assembled battery cabinets.

Skills

Languages: Python, C++, C, Go, Perl, Java, SQL, MATLAB/Simulink, VHDL
AI and vision: Deep reinforcement learning, PyTorch, OpenCV, point clouds, data analysis, filtering, numerical simulation
Robotics and embedded: ROS/RViz, Raspberry Pi, Arduino, IoT, Android Studio, 3D CAD (SolidWorks, CATIA V5)
Automation and DevOps: IEC 61499, CODESYS, SoMachine, Docker, Kubernetes, AWS, Git, Makefile, LaTeX
Spoken languages: English, French, Russian, Uzbek (C1 or native); German (A2)

Education

2020–2024: PhD, Université Côte d’Azur (I3S / INPHYNI)
2017–2020: MS in electrical engineering, INSA Lyon. Specialization in image and signal processing.
2015–2017: Preparatory courses, section SCAN (sciences in English), INSA Lyon
Online: Udacity Robotics (SLAM, deep RL on a robotic arm, Gazebo pick-and-place) and Machine Learning (AWS-sponsored). LabVIEW Associate Developer.
Outside work: Chess (national medalist, Elo 1969), karate (black belt), judo, swimming, ski. INSA Lyon robotics club Clubelek, French Cup of Robotics 2017–2018.