Memoji

Paul Chu

About Me

I'm a recent graduate with an MSc in Applied Mathematics and Computer Science from ENS Paris-Saclay, originally from Paris, France. I'm trilingual in French 🇫🇷, Chinese 🇨🇳, and English 🇺🇸, and currently in the U.S. on a J-1 visa valid until April 2026.

Currently based in Palo Alto and working as an AI Engineer, I'm passionate about soccer, going to the gym, and creating YouTube vlogs. I'm actively looking for an internship position as an AI/ML Engineer starting in November 2025, and I'm available to start immediately.

My research focuses on applications of fine-tuning LLM embeddings, and I've co-authored papers at Stanford University. I regularly participate in hackathons around San Francisco, love building things from scratch, and am always eager to learn more.

Academic Journey

ISEP — BSc & MEng in EE/CS

Five-year integrated engineering program (Paris)

  • • “Grande École d'Ingénieurs” engineering degree
  • First-Class Honours ranked Top 5/220 in my cohort
  • Two-year Preparatory Classes (Maths & Physics)Lycée Stanislas (CPGE), intensive track

Stanford University — Master's Thesis

Pilanci Lab (EE Department), supervised by Prof. Mert Pilanci

  • • Awarded a $15k Stanford Overseas Research Grant.
  • • Selected as 1 of 2 from 220+ students for the research program.
  • • Co-authored my first research paper about “Randomized Geometric Algebra Methods for Convex Neural Networks”
  • • Created video contents: Stanford vlogs documenting my adventure during the year
Watch the video

Stanford Medicine — Research Assistant

Nguyen Lab, supervised by Dr. Mindie H. Nguyen

  • • Hepatology research across HBV, HCV, NAFLD, liver cancer & transplant — from epidemiology to clinical trials
  • Global health collaborations (e.g. China, Taiwan, Japan, Spain)
  • • Focus on health disparities & ethnicity-related risk factors and outcomes
  • • Embedded in a small 3-Stanford-engineer data/ML team (with Mattia and Sovann) alongside doctors & researchers

AIIB — Summer SWE (Data Team), Beijing

First professional experience in China; data platform & analytics work

  • Asian Infrastructure Investment Bank (AIIB) is a multilateral development bank (HQ: Beijing) financing sovereign & non-sovereign infrastructure; mission: sustainable infrastructure to improve social & economic outcomes
  • Highly selective intake: <0.01% acceptance

ENS Paris-Saclay — MSc (MVA)

MVA (Mathematics, Vision, Learning): elite research Master’s (est. 1996), led by ENS Paris-Saclay; co-accredited with Institut Polytechnique de Paris, with École des Ponts & Université Paris Cité. Bridges rigorous mathematics with vision/signal and machine learning, tightly linked to labs & industry

  • • Often cited among top programs in Applied Math/AI
  • • Core: optimization, probability/statistics, ML/DL, computer vision, NLP, generative models, RL
  • • Outcomes: 250+ students/year; strong research & industry placements; annual Internship Forum (companies ↔ labs)

EF — AI Startup School, Paris

Selective program (~40 participants) for technical talent to learn directly from founders how to launch and grow AI startups. Five Monday evening “how-to” sessions + a 48-hour builders weekend finale

  • • Europe’s hands-on school for starting an AI company; candid, founder-led sessions
  • • Learned from French unicorn founders & industry leaders (e.g., Maki, Argil)
  • • Builders weekend: ship in 48 hours and demo outcomes

Y Combinator — AI Startup School, SF

Invited among ~2,000 CS students globally; learned from world-class founders/leaders (e.g., Andrew Ng, Elon Musk, Sam Altman, Satya Nadella, …)

  • • Main stage: Sam Altman, Satya Nadella, Elon Musk (Zoom), Andrew Ng, Andrej Karpathy, Fei-Fei Li, John Jumper, François Chollet, Aravind Srinivas, Jared Kaplan
  • • Breakouts: Dylan Field, Aaron Levie, Suhail Doshi, Varun Mohan, Jake Heller, Tom Blomfield, Sriram Krishnan, Pete Koomen, and more
  • • Format: keynotes, fireside chats, breakout tracks, community after-parties

Righ — Software Engineer, Palo Alto

AI team — LLM multimodal projects

  • • Founded in 2024, Righ brings together a seasoned team from leading technology companies. Over 80% of the team is dedicated to software and hardware development, with deep AI/ML expertise.

Professional Journey

Experience

AI Software Engineer

Full-time

Righ

Jun 2025 - Oct 2025

Worked on a multimodal product recommender (image + text + behavior); shipped offline eval +21% precision, +13% TPs; link filter: −12% low-quality at ≥82% precision.

Python, PyTorch, Grounding DINO, Real-ESRGAN, TF-IDF, Vector Search, YAML

AI DevRel Engineer

Part-time

Mistral AI

Jan 2025 - Jun 2025

Built and deployed Systempromptstral, a scalable web app (Next.js + TypeScript) for system prompt optimization with dynamic few-shot prompting and similarity-based evaluation.

Next.js, TypeScript, Python, Mistral AI API, LLMs

Summer Software Engineer

Internship

Asian Infrastructure Investment Bank

Jul 2024 - Sep 2024

Led automation of data extraction from the World Bank Group, creating a Project Insights Dashboard (PID) database and developing ChatPID, an AI-powered web interface for personalized project recommendations.

Python, LLMs, Data Visualization, Power BI, Web Development, Statistics

Research Assistant

Part-time

Stanford Medicine

Mar 2024 - Jul 2024

Streamlined meta-analyses processes, improving information retrieval for researchers. Developed and managed a system to store and organize information extracted from research papers, creating a dataset for efficient case study analysis.

Python, Data Analysis, Research, Database Management, Statistics

R e f e r e n c e L e t t e r
R e f e r e n c e L e t t e r

Student Researcher

Full-time

Stanford (Electrical Engineering Department)

Sep 2023 - Jul 2024

Ran large-scale transfer-learning experiments with GPT-4/BERT embeddings in PyTorch. Compared results from convex optimization vs fine-tuning on IMDb, Amazon Polarity, GLUE → +3.2% avg accuracy, −40% variance across seeds.Co-Authored a reproducible benchmark harness (CIs, seed sensitivity, data leakage checks); released clean code, reports.

Research, LLMs, Python, PytTorch, Benchmark, Convex Optimization

R e f e r e n c e L e t t e r
R e f e r e n c e L e t t e r

Data Analyst

Full-time

askR.ai

Mar 2024 - Jul 2024

Conducted cohort analysis to identify user engagement and retention trends, using SQL and Python (pandas). I built dynamic dashboards using Tableau and Power BI to visualize retention patterns and support decision-making. Then applied statistical modeling techniques in R to assess customer lifecycle and churn risk

Data Analysis, Python, Power BI, Tableau, R, Database Management, MySQL

Projects

Systemprompstral

Systemprompstral

Systemprompstral helps you refine system prompts using Mistral Large through structured training, testing, and few-shot workflows. Ideal for tasks like code generation, classification, and more.

React TypeScript WebSocket Mistral Large model
SkyTrip

SkyTrip

SkyTrip is a smart weather-aware flight planner that combines WindBorne weather data with Google Flights pricing to recommend destinations. Our algorithm scores destinations based on weather comfort (30%), flight price (40%), and travel duration (30%), helping you find the perfect trip that matches your climate preferences and budget.

Next.js TypeScript WindBorne API SerpApi
Synapse Travel Plan

Synapse Travel Plan

Synapse Travel Plan is an LLM-powered tool that creates personalized travel itineraries from user preferences, budget, and dates, with activity, hotel, and restaurant recommendations plus cost and weather insights

NextJS Asyncio Pydantic FastAPI
Guess Who's Robot

Guess Who's Robot

"LeCopain" is a friendly robot capable of playing the classic board game "Guess Who?" by combining natural language processing, speech interfaces, and robotic control into a seamless, interactive experience

Text-to-Speech Speech-to-Text Mistral AI model ACT policy
Magic Story Box

Magic Story Box

Our team developed Magic Story Box, an AI-powered interactive storytelling toy that lets children co-create stories with their favorite characters, entirely on-device. Using Whisper Tiny (speech-to-text), a fine-tuned Llama 3.2 3B model (story generation), and XTTS-v2 (voice cloning of Santa 🎅), we created a magical, privacy-safe toy experience without cloud dependency.

Whisper Tiny model Llama-3.2 model Text-to-Speech GPT-4o model
Open Source
Mistral DOOM

Mistral DOOM

We finetuned Mistral-7B model to make it capable of playing DOOM through visual input alone, utilizing the ViZDoom engine, a prominent tool in visual reinforcement learning. We transformed each game frame into a 32x64 grid, representing game elements in each cell, enabling the LLM to interpret the game via ASCII text.

Mistral-7B model C++ ASCII LLMs
Page 1 / 2

Media & Awards

Research & Publications

Research thumbnail
Research Stanford University

Randomized Geometric Algebra Methods for Convex Neural Networks

Co-authored research paper introducing randomized algorithms to Clifford's Geometric Algebra, generalizing randomized linear algebra to hypercomplex vector spaces. Key application area is fine-tuning LLM embeddings with comparative analysis on robustness of transfer learning.

View Paper
Fireworks.ai blog cover
Technical blog Mistral AI hackathon in SF

Technical article on Mistral DOOM with Fireworks.ai

Doomed to Code: How we Teamed Up with Fireworks AI at MistralAI Hackathon to Conquer the Shores of Hell

View Article
Hugging Face blog cover
Technical blog HF (LeRobot) hackathon in Paris

Technical article on our "Guess Who" robot with Hugging Face

Building an Interactive Guess Who? Robot with Mistral AI and SO-100 Arms

View Article

Hackathons & Projects

April 2024

Mistral AI Hackathon in SF

San Francisco • Mistral AI Hackathon • Winner

A 24-hour hackathon organized by Mistral AI, in partnership with Cerebral Valley and Shack15, bringing together builders to create AI-powered projects using the Mistral API and models. On March 23rd, 2024, I teamed up with Bhav, an ML researcher from San Francisco, and Umut, an ML engineer from Turkey. Together, we built a "Game Copilot", an AI agent designed to play and demonstrate in-game actions using the classic 1993 video game Doom. Our team leveraged Mistral-7B and Fireworks AI for rapid iteration and last-minute deployment within the intense 24-hour hackathon timeline. We won the firt prize and received $10K+ in cash prizes and credits.

View the Demo
November 2024

Consumer Edge Meta hackathon in Paris

Paris • Meta Hackathon • Winner

The Magic Story Box, the next generation of AI-powered toys that brings storytelling to life through dynamic interaction. Tomorrow's kids won't just hear stories, they'll actively shape them. Whether they want to explore tales with their favorite characters from cartoons or cultural icons, the Magic Story Box adapts to their curiosity. With my team (Ivan, Ali, Thomas, and Karim and I), we won the second prize and received $2,5K+ in cash prizes and credits.

View the Demo
January 2025

Mistral AI Game Jam Hackathon in Paris

Paris • Mistral AI Hackathon • Finalist

An in-person AI Game Jam in Paris (PhotoRoom), Jan 24–26, where ~100 participants build a playable game in 48 hours using Mistral AI models. Mentorship and resources from Mistral, Hugging Face, Scaleway, ElevenLabs, Nebius, and Balderton, with $8K+ in prizes and two awards: Jury Prize and Community Choice (Twitch).

View the Demo
March 2025

LeRobot hackathon in Paris

Paris • HF LeRobot hackathon • Winner

Our team: Alexandre, Arthur, Thomas and I, won 2nd place at the Mistral AI Robotic Hackathon, which challenged participants to combine Mistral's open‑source LLMs with SO‑100 robotic arms to build interactive real‑world systems. We built "LeCopain", a robot capable of playing the board game Guess Who? showcased how robotics, LLMs, and speech systems can merge to create engaging, educational, and assistive experiences We won the second prize and received $2,5K+ in cash prizes and credits.

View the Demo
August 2025

YC Agentic hackathon in SF

SF • Hackathon

The AI Coding Agents Hackathon at Y Combinator (Aug 10, 2025) gathered top AI developers to build next-gen coding agents. Teams created tools that generate, adapt, or run code based on human or autonomous inputs. Tracks included web/mobile apps, game creation, reactive agents, and evolving software. Winners earned YC interviews, GPT-5 access, and $200+ in perks from Morph, Freestyle, and Anthropic.

View the Demo

Contact

I'm always open to opportunities in AI/ML engineering, internships, or collaboration. Feel free to reach out directly: