First Months as Solutions Architect

This roadmap reflects my initiative for growing NVIDIA’s presence in scientific ML and AI research communities. I’d love to iterate on this with the team.

Project Draft

Objectives

• Support research labs in adopting deep learning and accelerated computing.
• Act as a technical bridge between NVIDIA’s products and the scientific ML community.
• Promote adoption of PhysicsNeMo, CUDA, and DLI via training and collaboration.

First 3 Months – Integration & Engagement

Lab Collaborations

• Initiate discussions with key labs where I have prior academic or research ties:
  • HPC@Maths Team of the Applied Mathematics lab of École polytechnique.
  • Maison de la Simulation Joint laboratory between CEA, CNRS, Université Paris-Saclay.
  • SERMA Reactor physics lab of CEA.
  • SGLS Software engineering for simulation lab of CEA.
  • LJLL Applied Mathematics lab of Sorbonne Université.
• Identify use cases for scientific ML (surrogate modeling, PINNs, inference).
• Connect PhD students and researchers with NVIDIA DLI certified training programs.
• Share my experience as a CUDA DLI trainee (during my PhD in 2021).
• Explore internship proposals and lightweight collaborative prototypes.
• Grant cloud computing credits.

Summer School Engagement

• Target top summer/winter schools not yet covered by NVIDIA:
  • The Gray Scott HPC summer school: 1 week CPU and 1 week GPU training
  • CEA–INRIA–EDF HPC School: The topic of the school 2025 was Solving partial differential equations in fields physics faster with physics-based machine learning
  • Hi! Paris Summer School on AI
• Propose NVIDIA sponsorship and/or send trainers to deliver DLI modules.

Conference Sponsorship & Outreach

• Identify and propose sponsorship opportunities in relevant applied math/AI conferences:
  • Mathematical Summer Center for Advanced Research in Scientific Computing CEMRACS
  • European Conference on Numerical Mathematics and Advanced Applications ENUMATH
  • National Numerical Analysis Congress CANUM
  • International Conference on the Physics of Reactors Physor 2026 in Torino
• Deliver demos, PhysicsNeMo workshops, or DLI technical talks.
• Full list: SMAI Event Calendar

Webinars & Enablement

• Host a public or academic webinar (e.g. “Getting Started with PhysicsNeMo for Scientific ML”).
• Write a technical blog post reflecting early feedback from researchers.
• Start a prototype PINN demo (e.g. solving the heat equation with PhysicsNeMo).

12-Month Milestone – Plasma Physics Prototype

Deep Learning for Plasma Simulation

• Identify and contact plasma simulation labs:
  • ITER, ENSTA, INRIA's CASTOR and TONUS teams.
• Discuss use cases for PINNs, operator learning, or surrogate modeling.
• Build a PhysicsNeMo-based prototype for a simplified plasma PDE (e.g. magnetohydrodynamics).
• Share outcomes in technical workshops.

Targeted Deliverables

• 3+ academic lab collaborations initiated
• 1 NVIDIA training or workshop delivered
• 1 summer school with NVIDIA content integrated
• 1 technical prototype developed and shared
• Technical blog post or internal note published

Thanks for reading! Go back for more notes.