Erwin POUSSI

Erwin POUSSI

Aeronautics & Astronautics @ Stanford | Research Assistant (MSL & Qiu Lab)

Stanford University

Biography

I am currently pursuing an M.S. in Aeronautics & Astronautics at Stanford University (2025–2027), where my work focuses on autonomous systems at the intersection of perception, decision-making, and learning.

At Stanford, I am a Research Assistant in Prof. Xiaojie Qiu’s lab, contributing to Pantheon-CLI, an open-source, LLM-powered agent framework for scientific analysis at the Stanford School of Medicine. My research centers on multi-omics workflows and agent-based methods for autonomous gene panel design. I am currently fine-tuning and continually pretraining Qwen3 LLMs to power Pantheon’s self-evolving agent backbone, leveraging reinforcement learning from verifiable rewards (RLVR) and tool-augmented inference to build domain-adaptive, evolvable agents for autonomous genomic reasoning.

I also conduct research in the Multi-Robot Systems Lab with Prof. Mac Schwager, where I work on instruction-conditioned drone navigation, combining language-based goal inference with autonomous control. In parallel, I develop structure-from-motion pipelines for 3D reconstruction using drone-based photogrammetry.

Prior to Stanford, I completed the Cycle Ingénieur Polytechnicien (Master’s-level degree) at École Polytechnique (France), with a minor in Entrepreneurship, focusing on control, optimization, statistics, and computational methods.

Outside of research, I enjoy sports—particularly strength training—and previously practiced track and handball. I highly value time spent with family, friends, and community service initiatives, which play an important role in my personal and professional life.

Interests
  • Autonomous Robotics
  • Reinforcement Learning
  • Computer Vision
  • Decision-Making under Uncertainty
  • Control & Optimization
Education
  • MS in Aeronautics & Astronautics, 2025–2027

    Stanford University

  • Master’s-level program in Mechanical Engineering (minor in Entrepreneurship), 2020–2025

    École Polytechnique

Selected Work

.js-id-‘*’
GAIT-Manip: Whole-Body Robot Control for Loco-Manipulation
Whole-body control of G1 and Spot robots for loco-manipulation, using GAIT scene reasoning.
GAIT: Graph-Memory Agent with Interrupt-Triggered Reasoning
Leverage online semantic scene graph construction to build a persistent memory for vision-language-action models, enabling open-vocabulary navigation.
Growing a Virtual Embryo with 3D Convolutional Models
Training foundational models to reproduce embryogenesis — predicting both cell-level gene expression dynamics and 3D morphogenesis.
Language-Steered Drones
End-to-end robot learning pipeline for language-guided drone navigation — from MPC expert demonstrations to sim-to-real deployment via behavioral cloning and DAgger.
Pantheon: LLM-Based Bio Assistant
A multi-agent LLM system that automates gene panel selection for large-scale spatial transcriptomics.
SelectSmart: Autonomous Gene Panel Design
A reinforcement-learning framework for adaptive gene panel selection that preserves single-cell structure while minimizing panel size.
Physics-Informed Random Forest Surrogate Modeling for Parachute Fabric Permeability
A surrogate modeling and optimization framework for pore-scale permeability inference in supersonic parachute fabrics.
Autonomous Spacecraft Docking
Undergraduate research project on autonomous spacecraft docking conducted in collaboration with Nyx Exploration Company.

Contact