Robotics & Controls Portfolio

Clement Dumeril.

Robotics and controls engineer working on model-based control and physics-informed learning for physical systems.

MEng Mechanical Engineering student at UC Berkeley (2026-2027). My background is in industrial mechanics from Arts et Metiers - CAD design, finite element analysis, multiphysics simulation, machining, metrology and vibration theory.

Across projects in industrial gearbox design, manufacturing studies, acoustic and vibratory systems, robotic cells and simulation-driven engineering missions, I connect physical modeling with practical technical execution - the foundation the model-based control and physics-informed learning work above builds on.

See featured work Browse project vault

Advanced control and robotics
AI-assisted engineering workflows
CAD / FEA / multiphysics simulation
Industrial mechanics
Manufacturing / machining / metrology
Vibration and experimental systems

Academic context

Arts et Metiers Industrial mechanics

UC Berkeley Mechanical Engineering - Robotics

01. Featured Missions and Applied Work

Cryogenic container temperature simulation

Cryogenic Container Mission

CAD (CATIA V5) - CFD (COMSOL) - Management (Team of 3)

Managed a team of three Arts et Metiers engineering students on a consulting mission for Airflow, focused on the development of a liquid nitrogen-based refrigerated container. The project aimed to maintain -3°C for 60 hours by optimizing the internal piping architecture. Delivered technical design recommendations and conducted a quality and regulatory analysis for the transport of perishable food products, supporting the future industrialization of the prototype.

$15k mission value
4 months of study

Ministry of the Interior

Innovation - CAD (Fusion) - Prototyping - Testing

Led a 5-member engineering team on a law-enforcement innovation program and worked on a vehicle immobilization device improving operational safety and control.

Management of a 5-engineer team
Best Engineering Study Award - Alten Prize 2025
LinkedIn post - Study highlights
$60k+ mission value
7 benchmarked innovation ideas
150+ fabrication hours

Silent Fan Simulation Package

CFD - Modal analysis - Dynamic impact - Technical reporting

As part of a project carried out with a friend for a company, I handled the simulation work on a low-noise fan concept inspired by biomimetic blades. The mission covered the CFD study of the fan itself, the modal behavior of the PETG-CF housing and the dynamic shock analysis of the assembly in order to document flow quality, structural modes and impact robustness in one coherent engineering package.

3 studies grouped into one simulation package
CFD, modal and shock analyses on the same fan system
Industrial side project delivered with technical reports and visuals

CFD report Modal report Shock report

Type IV hydrogen bottle mission setup with robotic bench

Type IV hydrogen bottle prototype detail on test bench

Type IV hydrogen bottle prototype being carried

Type IV Hydrogen Bottle Prediction

PyTorch - Abaqus - CalculiX - Genetic algorithm

This project focused on correcting and restructuring a genetic algorithm used to predict rupture in a Type IV hydrogen tank. The workflow combined machine-learning logic with a DOE campaign generated from CalculiX simulations, then used that dataset to recalibrate the prediction pipeline and study failure trends across the vessel design space.

Type IV composite rupture prediction study
DOE dataset generated from CalculiX simulation runs
Final reports available in French and English

Final report FR Final report EN

Dream Extraction Lab interface for fiche idee extraction comparison

PJT - Idea Sheet Extraction Study

Fine-tuning - Hugging Face - VLM - DODO

LCPI and Arts et Metiers project partners logos

One-year research support project carried out with LCPI and Arts et Metiers to help a doctoral researcher study how engineering data can be extracted from both an idea-sheet diagram and its accompanying text in order to reconstruct a usable 3D model. We fine-tuned a DODO-based vision-language model on a reworked ABC library through Hugging Face, then compared that pipeline against several public large models to measure the impact of engineering-oriented prompting on extraction quality.

1 year of work contributing to an LCPI doctoral thesis
DODO fine-tuning built on a reworked ABC library
Public LLMs and VLMs benchmarked under engineering prompt design

Open extraction lab Methodology Scoring guide

02. Relevant Experience

Sales Director / Project Manager @ AMJE

Arts et Metiers Junior Etude, Paris. Progression from project delivery to sales leadership.

Feb 2025 - Present

Sales Director

June 2025 - Present

Lead commercial development for AMJE in Paris across engineering consulting opportunities.
Drive client acquisition, proposal framing and continuity between business development and project execution.
Built on delivery experience to strengthen the commercial positioning of AMJE projects.

Project Manager

Feb 2025 - June 2025

Managed engineering consulting studies on site with responsibility for delivery quality and client communication.
Coordinated project scope, milestones and team execution before moving into the sales director role.

Project Manager - Engineering Consulting @ Ministry of the Interior

Part-time hybrid mission in Paris on a strategic law-enforcement innovation program.

June 2025 - Present

Lead a 5-member engineering team on a strategic innovation project commissioned by Commissaire General Thierry Callies, under the supervision of Gaelle Casagrande and Hammama Lafane.
Invented a vehicle immobilization device for law enforcement that can stop vehicles without puncturing tires, improving operational safety and control.
Project awarded the Alten Prize for Best Engineering Study in France in 2025 and renewed for a second phase covering automation, prototype scaling and preparation for full-scale operational testing.

Project Manager - Engineering Consulting @ AIR FLOW

Part-time mission focused on thermal optimization and risk analysis for railway cooling systems.

June 2025 - Present

Lead a multidisciplinary engineering team on thermal optimization and risk analysis for the integration of liquid nitrogen cooling in a mobile railway environment.
Conducted thermal studies, heat-exchange geometry optimization and structured risk assessment across design and operational phases.
Delivered technical recommendations to improve safety, performance and regulatory compliance.

Engineering Intern @ Groupe Techwave

AI integration mission in an industrial environment focused on mechanics and electronics.

June 2025 - July 2025

Acted as the interface between OPEO AI consultancy and Techwave engineering teams.
Researched and evaluated AI models adapted to Techwave requirements and industrial use cases.
Contributed to securing EUR 30k in BPI funding through the Accelerateur PME program.

03. Project Vault

Secondary projects, experiments and side builds. The cards below work like a dynamic project slider: pick one card to update the detail panel.

Robotics / Vision / Physics-Informed Neural Networks / Webots

UR5e Vision + PINN Inverse Kinematics

UC Berkeley robotics case study connecting a Webots vision dataset, a trained VGG16 coordinate regressor, a physics-informed neural inverse-kinematics model, differentiable UR5e forward kinematics and an IKPY benchmark inside the Webots controller loop.

VGG16 vision
PyTorch PINN
UR5e FK
Webots benchmark

Open scientific case study View dataset plot View timing plot

04. Archive

Supporting files, reports and dedicated viewers. The archive is now ordered from the most recent work to the earlier projects so the progression stays readable at a glance.

2026

Recent robotics, simulation, browser tooling and technical delivery, from vision-driven control to CFD, impact and translation workflows.

2025

Recent applied research, simulation and automation work.

2024

Manufacturing projects and applied sustainable design work.

2023

Foundational mechanics, acoustics and early systems modeling.

Robotics / Vision / Physics-Informed Neural Networks / Webots

2026

UR5e Vision + PINN Inverse Kinematics

Scientific robotics case study built from real project artefacts: Webots vision dataset, VGG16 coordinate regression, physics-informed UR5e inverse kinematics, differentiable forward kinematics and IKPY timing benchmark.

VGG16 vision
PyTorch PINN
UR5e FK
Benchmarking

Open scientific case study View dataset plot View timing plot

UR5e Vision + PINN Inverse Kinematics archive preview

05. Contact

Direct access to projects, reports and contact.

The portfolio is structured to let recruiters scan the main engineering work first, then open the supporting documents they need without friction.

GitHub Get in touch Download resume