Built an AlphaZero implementation from scratch in JAX/Flax, with parallelized self-play that sped up training by 2x on a compute cluster

Analyzing AlphaZero training dynamics on solved Chinese Checkers, using game-theoretic ground truth to diagnose training anomalies and evaluate model accuracy

Self-taught formal logic and proof material each week to a sufficient level to present solutions and field student questions live for a class section of ~100 students

Marked quizzes, exams, and mini-essays; held office hours; and supported students through course forums and email

Authored weekly quiz and final exam questions through multiple rounds of instructor feedback, developing an eye for questions that test conceptual understanding over memorization, calibrated for a non-specialist audience

Built a Python script using the Canvas REST API to bulk-collect peer review scores that Canvas provided no native export for, navigating several interconnected API objects to reconstruct per-student data

Designed a Python + Google Sheets pipeline to automate peer review grading: the script pulled raw scores via API, the sheet computed weighted averages, flagged anomalies, and generated a CSV for bulk Canvas grade upload

Worked on an entity linking research project supervised by Professor Anoop Sarkar and his PhD student Hassan Shavarani

Read entity linking papers and reproduced their results using their public repositories and models

Conducted follow-up experiments on the reproduced entity linking methods

Wrote and published a paper (Unified Examination of Entity Linking in Absence of Candidate Sets) in the NAACL 2024 Conference

Delivered case management software to clients by coordinating with the product team, operations team, and the clients

Collaboration in agile development teams done using software such as JIRA, Teams, and GitHub

Built Redmane Canada's first live interface using the Azure ecosystem and REST APIs, saving the client thousands of hours per year of manual data transfer between our system and theirs

Used C#, .NET, and related libraries to implement client requirements for their software solution

Used Transact-SQL scripts and queries into the database to automate development work, investigate issues, and solve identified problems

Led design sessions with clients to determine their organizational needs and how we can solve them

Trained clients to use our software, ensuring a smooth transition for their organization to our solution

Performed maintenance on production systems through regular monitoring

Coordinated duties within a team in a fast-paced and quickly changing environment for long (sometimes 9+ hour) shifts

Handled rapidly growing dish piles while prioritizing tasks based on height of pile, line cooks' needs, and workplace maintenance

Hauled and organized flats

Assisted in setting up the lighting for the studio and the construction of the set

Procedural Content Generation via ML: Studied generative methods for game content including Markov chains, CNNs, VAEs, GANs, diffusion models, and reinforcement learning-based generation

Reinforcement Learning: Studied MDPs, policy iteration, Monte Carlo methods, temporal-difference learning (SARSA, Q-learning), function approximation, and deep RL

Experimental Mobile Robotics: Studied differential drive kinematics, computer vision, sensor fusion, path planning, and localization using ROS and OpenCV

Neurosymbolic Programming: Studied program synthesis methods including top-down and bottom-up search, neural-guided search, stochastic methods, latent space learning, and transformer-based synthesis

Awarded President's and Dean's Honour Roll and scholarships for excellence in grades

Created complex SQL queries, utilizing knowledge of relational algebra and calculus

Parallel Programming: Worked with multi-threaded applications, race conditions, critical sections, mutual exclusion locks, and semaphores

Learned programming paradigms - functional, concurrent, and distributed - with Haskell, Go, and Erlang

Used TensorFlow and MATLAB to build Convolutional Neural Networks and traditional edge+point detection systems

Wrote a Quantum Program Simulator to run any quantum program (sequence of gates) the user writes

Studied various NLP techniques, such as RNNs, LSTM, ELMo Embeddings, BERT, Word2vec, etc

Hand wrote machine learning algorithms such as entropy-minimizing random forests, k-modes clustering, and LOF anomaly detection

Learned CPU architecture, wrote in Assembly, and developed various OS components such as memory management

Studied internet protocols from each layer (eg: DNS, TCP, CSMA/CD, etc) and other networking topics such as Data Centers and VoIP

Worked with ROS/rospy - nodes, topics, services, messages, and rosbags.

Performed localization using odometry with the robot's wheel encoders.

Collected the generated odometry data; used it to visualize the robot's path.

Used OpenCV for computer vision - color detection, apriltag detection, and lane detection.

Implemented PID controllers to perform lane following.

Bonus: Built a PyBullet Simulation to compare PID and MPC controllers balancing balls on a plate. Report.

Developed in C++ using the graphics library Vulkan

Implemented Ray Tracing principles in a compute shader, which then renders the final image directly onto the swapchain's images

Utilized GPU-accelerated computation to calculate each pixel's data in parallel

Learned about blockchains by developing a blockchain networking protocol based on Bitcoin's original whitepaper

Major features include networking capability, proof of work, longest chain, and mining functions

Utilized multiprocessing and concurrency to develop a simulated socket interface, allowing one machine to simulate multiple nodes over the internet

Used TCP sockets at the application layer to allow a single machine to act as a node for this program on the internet

Extracted relevant information from big data (OpenStreetMap dataset) using Spark and Hadoop on SFU's compute cluster

Feature engineering with Pandas and Numpy, training and optimization with scitkit-learn (Random Forest, Pipelines, and Grid Search), visualization with Plotly

Trained a model that can explain 35% of a hotel's score knowing only its location

Written in Lua using the LOVE2D framework and Box2D for the physics engine

Created procedurally generated asteroids and asteroid fields using Perlin noise

Wrote a line-of-sight algorithm so the player can't see behind asteroids

Became a full-stack web developer - learned node.js/PostgreSQL for the server-side framework, and javascript/HTML/CSS for the client-side

Deployed with cloud services such as Amazon's AWS (S3 for image hosting; SES for confirmation emails), and Heroku (Platform for node.js app hosting; Postgres for database hosting)

Designed the database's structure for the website and created a Entity-Relationship Diagram to visualize it

Developed in the Android Studio environment and released to the Google Play Store

Designed the application using Java for the model and controller, XML for the view, and SQLite for the item database