Free iPhone & Android app that turns shared interests into real meetups.
PLAYER 1 · READY
MATIN MOBINI
CS STUDENT @ UOTTAWA // BUILDS AI/ML, WEB & ANDROID APPS
- 3.9/4GPA
- 12PROJECTS
- ∞PASSION
MATIN MOBINI
HI! I'M MATIN. I BUILD AI/ML, WEB & ANDROID APPS. I STUDY COMPUTER SCIENCE AT UOTTAWA, CONCENTRATING IN AI & ML. FUN FACT: I'M NOT PRESSING ANY BUTTONS. A NEURAL NET IS PLAYING ME. IF I FALL IN A PIT, THAT'S NOT A BUG. IT'S A LEARNING OPPORTUNITY. I TRAINED FOR HUNDREDS OF GENERATIONS AND STILL FORGET TO JUMP SOMETIMES. THIS HEADSET IS 100% FOR THE AESTHETIC. MY CODE WORKS ON THE FIRST TRY. MOSTLY. SOMETIMES. ONCE. SCROLL DOWN AND I'LL JUMP INTO THE NEXT WORLD. NO PRESSURE. THE ENEMIES HERE TALK TRASH. IGNORE THEM. WANT THE CONTROLLER? OPEN THE AI BRAIN AND PRESS PLAY IT YOURSELF.
WORLD 2
ABOUT
WHO I AM, WHAT I'M GOOD AT AND WHAT I'M LEARNING.
- NAME
- MATIN MOBINI
- CLASS
- HONOURS CS · AI & ML
- HOME
- OTTAWA, CANADA
- LIKES
- GAMING, BASKETBALL, GYM
A passionate developer with a knack for problem solving and teamwork. Honours Computer Science at the University of Ottawa (GPA 3.9/4.0), nine-time Dean's Honour List and a Merit Scholarship. Levels up in Artificial Intelligence & Machine Learning, Web Development and Android Development.
ALSO SPEAKS
- PYTORCH
- SCIKIT-LEARN
- PANDAS
- NUMPY
- HUGGING FACE
- OPENAI API
- RAG
- PROMPT ENGINEERING
- STREAMLIT
- FASTAPI
- DOCKER
- JUPYTER
- C++
SKILLS
- PYTHON
- JAVA
- JAVASCRIPT / TS
- SQL
- AI / ML
- REACT + NODE
- ANDROID
ACTIVE QUESTS
- RALLY A social app that brings people together through shared interests.
- MENTAL HEALTH ASSISTANT A context-aware RAG phone app that reads your smartwatch data to track your health and help you through stressful times, built on real therapist methods.
- CROPPILOT An autonomous robot for crop detection, precision harvesting and storage organization.
WORLD 3
EXPERIENCE
WHERE I'VE WORKED AND STUDIED.
-
- + Lead development of AI and robotics software in Python, delivering substantial enhancements to core intelligent systems and autonomous agent capabilities
- + Design and implement key features spanning Memory Management, Perception, Recognition and Action Execution, integrating sensory inputs, internal state and agent behaviour within robotic simulations
- + Translate AI research concepts into functional software and evaluate system behaviour through simulation, structured experimentation, root-cause analysis and regression testing
- Python
- Robotics
- Autonomous Agents
- Simulation
- Regression Testing
-
- + Engineered a scalable, multi-government-branch automated Case Management Tool, expanding across branches with the vision to supersede legacy systems and unify workflows and information storage organization-wide
- + Integrated an intelligent local decision assistant into the tool that analyzes proprietary, complex JSON datasets and recommends optimal solutions in real time, while answering inquiries about protected or online data
- + Worked independently on multiple concurrent projects with Python, SQL and data processing libraries in a hybrid government environment
- Python
- SQL
- Machine Learning
- Lead Developer
- Data Processing
-
- + Engineered a chemistry identification tool that runs 881 substructure tests on molecular data to generate binary fingerprints, helping chemists identify and understand undocumented compounds
- + Developed ChemPath, an app for visualizing and analyzing chemical synthesis pathways, used to inform officers about potentially hazardous molecules such as fentanyl and methamphetamine
- + Worked independently on multiple concurrent projects with Python, SQL and data processing libraries
- + Collaborated with chemists and data analysts so the tools met scientific and operational requirements
- Python
- SQL
- Data Science
- Cheminformatics
-
- + Implemented new features and enhancements to the company website: optimized user interfaces, refined navigation menus, interactive components
- + Contributed to server maintenance and backend development, including data collection and management with MySQL
- JavaScript
- TypeScript
- Node.js
- Java
- Python
- SQL
- MySQL
- GitLab
-
- + Concentration in Artificial Intelligence & Machine Learning
- + GPA 3.9 / 4.0, nine-time Dean's Honour List, Merit Scholarship
- + Vice President of HKFC, a uOttawa student club
- + Coursework: Machine Learning, Data Science, Data Structures & Algorithms, Probability & Statistics, OOP, Databases, Computer Architecture, Programming Paradigms
- Python
- Java
- SQL
- C++
- Racket
- Prolog
- 6THYOUR TEAM?SOFTWARE / AI & ML DEVINSERT COIN
WORLD 4
PROJECTS
MY CODING PROJECTS. CLICKTAP A CARTRIDGE FOR DETAILS.
Context-aware RAG therapist assistant that reads your smartwatch data.
PRIVATE REPO
Autonomous robot for crop detection, precision harvesting and storage.
PRIVATE REPO
This site. A neural net trained by neuroevolution plays every level.
Flags unfaithful sentences in news summaries, and explains why.
From-scratch PyTorch CNN that classifies brain MRI scans (85% test accuracy).
Next-day GOOGL return forecasting with a leakage-free backtest.
Prompt engineering that turns your diet profile into a personal recipe.
uOttawaHack 8: sports stats explained by an AI buddy with personality.
ML model that predicts house prices (Random Forest, RMSE 50,414).
Android app to run cycling club events and members.
SQL-powered hotel search, booking and reservation manager.
WORLD 5
NEURAL NET
THE AI THAT PLAYS THIS SITE. IT'S A REAL NETWORK, TRAINED BY EVOLUTION.
- GENERATION
- -
- FITNESS
- -
- PARAMS
- -
- EPISODES
- 0
- SKILLS
- 0/8
TRAINING FITNESS
BEST AVERAGE
SKILLS THE AI HAS SHOWN YOU 0/8
Each one unlocks the first time you see the runner do it.
HOW IT WORKS
The brain. A small neural network (a multilayer perceptron: 8 inputs, hidden layers of 10 and 8 neurons, 5 outputs, 223 weights) controls the runner. Every frame it reads 8 senses: distance to the next gap, gap width, distance and height of the nearest enemy, a coin above, a wall ahead, whether it's on the ground, and its vertical speed. Its 5 outputs decide how fast to run and whether to jump, double jump, dive or dash.
How it learned. Nobody hand-coded the moves. The weights were found offline with neuroevolution: a genetic algorithm ran a population of networks on seeded levels of all six themes, scored each one (distance, coins and stomps, minus a penalty for dying), kept the best, and bred the next generation with crossover and random mutation. The script is npm run train; the chart shows its best and average fitness per generation. The same physics code runs here and in training, so the site plays exactly the brain that was trained.
The growing map. The full network drives the runner the whole time. The map starts mostly hidden and reveals a pathway (the strongest weighted connections from the relevant senses to an action) the first time you see the runner use that behaviour. Yellow edges are positive weights, pink are negative, thicker means stronger; squares glow with live activation.
Your turn. Press PLAY IT YOURSELF to give the AI a break and take the controls.
Watch it learn. Press TRAIN IT YOURSELF to open the Training Lab: a fresh population of ghost runners starts from random weights and learns a level live, generation after generation, in your browser.
WORLD 6 · BOSS CASTLE
CONTINUE?
GET IN TOUCH BEFORE THE TIMER HITS ZERO. OPEN TO INTERNSHIPS / FULL-TIME / FREELANCE.