I build agentic AI systems that ship. I've taken enterprise bots from prompt to production at Botpress and built RAG pipelines over half a million support tickets at Retail Realm.
Architected an autonomous bot-building agent that generated 6-figure revenue growth by automating user onboarding and accelerating time-to-value for Enterprise customers.
Developed an LLM-powered orchestration layer that increased Day-1 user retention by allowing users to autonomously generate, test, and deploy production bots via natural language.
Engineered open-source integrations for SharePoint and MailerLite, achieving production-grade stability and high-reliability data ingestion for enterprise-scale unstructured data.
Led a strategic pilot program for Team Plan customers, translating direct feedback into core features that ensured 100% compliance with corporate accuracy standards prior to full automation.
AI Software Engineer (Intern)
05.2025 - 08.2025
Retail Realm
Developed an agentic chatbot utilizing the ReAct (Reasoning and Acting) framework, enabling the system to autonomously retrieve, reason, and respond to queries with reduced average support response times.
Deployed a production RAG system using Azure Databricks and PySpark to process 550,000+ historical support tickets, significantly reducing manual review time through automated semantic search.
Optimized inference performance by benchmarking and self-hosting open-source LLMs via vLLM on private virtual machines, resulting in reduced latency and API costs compared to third-party providers.
Implemented an automated ETL pipeline for historical ticket data, utilizing NLP techniques to improve the precision of contextual answer extraction from high-noise system logs.
Education
McGill University
2022 - 2026
BSc Honors Computer Science / Minor Statistics
Selected Coursework
Natural Language ProcessingArtificial IntelligenceMachine LearningReinforcement LearningTime Series AnalysisSoftware DesignApplied RegressionDatabasesData Structures and Algorithms
Reimplemented Rainbow DQN from scratch, integrating six core enhancements (Double Q-Learning, PER, Dueling/ Noisy Nets, n-step, C51) to achieve 153.9 average test reward on Seaquest. Executed ablation studies on Noisy Networks and Prioritized Experience Replay (PER), identifying the critical components required for stable reinforcement learning in sparse-reward environments.
PytorchNumpyGymnasium
Turing Poker Bot
Poker Agent
Developed a real-time decision engine using expectation calculations and opponent modeling, resulting in cash prizes in two competitive rounds of high-stakes play. Integrated a moving average RL concept to adapt strategies based on evolving opponent behaviors, maintaining positive expected value (EV) in dynamic environments.
Python
Digit Recognition with Convolutional Neural Network (CNN)
Python, Numpy, Pytorch, Pandas
Achieved 86%+ accuracy in recognizing handwritten digits Implemented techniques like batch normalization, data augmentation, and stochastic gradient descent to improve model performance and reduce overfitting
— Human Layer
What the resume leaves out.
— 02 - The Human
me & Bo, grad day
Hey, I'm Max.
This is the part that doesn't fit on one page. That's me on the left, next to my awesome girlfriend, Bo Lau.
When I was 10, my dad blocked me from the internet to stop me from playing games. I didn't ask him to unblock me, instead, I figured out how to bypass it myself. That instinct never really left me.
I ended up at McGill studying computer science right as AI was going mainstream, using it in my day-to-day. I got drawn to the practical side. What can actually be automated, and how do you use that to genuinely improve people's lives?
—Sneak-peek into my life
après ski · jay peak
last day in montreal
met eric shoji
piggyback through kyoto
harvard stadium · 2024
board game night
dig dynasty
hong kong at night
crashed in the library
convocation · mcgill
sus gala ii
arashiyama
boyfriend duties
ski · jay peak
we made it
guangzhou nights
before the walk
times square
sus gala · the crew
kimonos at the temple
fear of unemployment
ocean park · hk
après ski · jay peak
last day in montreal
met eric shoji
piggyback through kyoto
harvard stadium · 2024
board game night
dig dynasty
hong kong at night
crashed in the library
convocation · mcgill
sus gala ii
arashiyama
Since then I've shipped production AI agents at startups and late-stage companies, and I've learned something new from every person on every team I've been on. I'm currently looking for a new role in New York City where I can own what I build and keep learning fast.
Outside of work, I'm usually playing volleyball, snowboarding, or completely consumed by whatever hobby I've picked up that month.
—Hobbies
Volleyball
The sport I'll drop everything for. Open gym a few times a week, and I never need convincing to show up.
Cycling
My version of a long drive to clear my head. Going all the way out to Niagara from Montreal on a bike is the kind of stupid I love.
Climbing
Taught me that everyone has their own way of solving problems.
Travelling
New city, new food, new way of seeing things. I travel to eat and experience different cultures and scenery.
Snowboarding
Nothing beats renting a winter chalet (ideally with hot tub) with friends and going snowboarding the whole weekend.
Gaming
How I relax and stay connected with friends. Mostly competitive, because improving is half the fun.
Curious about right now
->The reliability bottleneck in AI.
->Applications of LLMs and their impact.
->Optimizing how I learn, not just what I learn
->When to keep a human in the loop
->How to stand out, when everything is being regressed to the mean
