Eric-Tuan Lê

Ai Research Scientist @Meta

London, GB
MOBILE NUMBERS
+91 *********19

Signup · Get unlimited contacts

WORK HISTORY

Mar 2024 — Present

Ai Research Scientist @Meta

View department →

London, GB

o GenAI 3D: Trained and evaluated flow-matching models for 3D shape editing/enhancement, learning time-conditioned velocity fields over SDF fields; improved geometric fidelity, edit controllability, and stability with few-step ODE sampling, classifier-free guidance, and ablations on path parameterizations.o Llama post-training (multimodal: text + image): Contributed to Llama 3 & 4 across three tracks(1) structured-image tuning/alignment to reliably parse and follow instructions over charts, tables, forms, and UI screenshots;(2) synthetic QA: scaled generation and filtering of instruction–answer pairs to broaden coverage and difficulty, improving instruction-following and factuality;(3) reasoning models with SFT → RL: supervised fine-tuning for policy initialization followed by RL-based preference optimization to strengthen step-by-step reasoning quality and answer consistency.

EDUCATION

2016 — 2017

ENS Paris-Saclay

Master of Research in Computer Vision & Machine Learning (Master MVA), Computer Vision and Machine Learning

2011 — 2013

Lycée Hoche

Preparatory classes for French "Grandes Ecoles", GPA: 4.00, Mathematics and Physics

2013 — 2017

CentraleSupélec

Master of Science in Applied Mathematics

2014 — 2017

ESCP Business School

Master In Management

2016 — 2016

Singapore Management University

Exchange Program

N/A

UCL

Doctor of Philosophy - PhD, Computer Vision & Graphics, Deep Learning

ABOUT ERIC-TUAN LÊ

I’m a Research Scientist at Meta AI focused on multimodal reasoning in Llama vision-language models. I work on SFT→RL post-training to improve step-by-step reasoning quality and answer consistency on text+image tasks. On a separate track, I’ve done post-training for structured images (charts, tables, forms, UI) and synthetic QA generation/filtering to broaden coverage. I previously worked on 3D generative modeling: flow matching over SDFs for 3D shape editing and enhancement, and earlier, SDS for 4D (3D + time).I earned my PhD at UCL’s Smart Geometry Processing Group under Iasonas Kokkinos and Niloy J. Mitra, specializing in deep learning, computer vision, and computer graphics, with a focus on 3D representation and reconstruction. Prior internships include Adobe Research (ICCV’21 + patent), Snap (CVPR’24 + patent), and Meta AI. I also review for CVPR, ICCV, ECCV, and NeurIPS.

This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.