z = 1.150r = 1

Selected work

Projects

Fig. II.1

Adaptive RL Ensemble Strategy

ensemblePPO·A2C·TD3SPY9 ETFs
Outcome:
Adaptive trading strategy that reduces portfolio volatility and improves risk-adjusted returns using an ensemble of reinforcement learning agents.
Tech Stack:
Python, PyTorch, PPO, A2C, TD3, Git
Key Constraints:
Adaptive allocation across varying market regimes, effective diversification to minimize drawdown, rigorous backtesting against SPY benchmark.

Developed at Queen's AI Club (QMIND), this project trains an ensemble of Reinforcement Learning agents (PPO, A2C, TD3) to rebalance a portfolio of 9 ETFs on a daily basis. Each agent learns inside a custom Gymnasium trading environment built on top of Stable-Baselines3, then their decisions are combined into a single adaptive allocation.

Rather than trusting a single backtest, I stress-tested the strategy across market regimes and modeled realistic 10 bps transaction costs. The results held up under scrutiny:

  • Risk-adjusted returns: an annualized Sharpe ratio of 1.14, achieved at roughly half the drawdown of a passive S&P 500 benchmark.
  • Statistical confidence: performance validated with 10,000-sample bootstrap confidence intervals, so the edge wasn't just a lucky run.
  • Adaptive diversification: the ensemble shifts weight across ETFs as conditions change, smoothing volatility instead of riding any single position.
Fig. II.2

Autonomous Multi-Agent Robotic Firefighting

Outcome:
Optimized autonomous robot deployment system for wildfire containment, achieving a 30% reduction in cluster convergence time.
Tech Stack:
MATLAB, Lloyd's Algorithm, K-Means Clustering, GIS Spatial Analysis Tools
Key Constraints:
Real-time dynamic fire hotspot data processing, accurate perimeter modeling, efficient multi-agent coordination.

Course project for APSC 200: Engineering Design and Practice. Course Grade: A+ | Role: Primary Technical Contributor

Driven by an interest in mathematical optimization, I developed a decentralized multi-agent system designed to tackle the unpredictable nature of wildfires. Moving away from rigid, centralized control, I implemented Lloyd’s Algorithm (i. e. K-Means Clustering) to allow a fleet of drones to "self-organize" based on real-time environmental data.

Multi-agent fire containment simulation
Drone agent trajectory plots

I built a custom simulation environment in MATLAB where I experimented with:

  • Dynamic Spatial Partitioning: Using Voronoi regions to ensure agents automatically distribute themselves to the most critical "hotspots" without human intervention.
  • Sensor Fusion Logic: Designing an "Observation Set" function that allows agents to weigh environmental density, effectively giving the swarm a collective "vision" of the fire's intensity.
  • Robustness & Scalability: Engineering the system to be decentralized so that the failure of one drone doesn't compromise the mission, as neighboring agents autonomously re-calculate and compensate for the gap.
Fig. II.3

Second Brain

Outcome:
A public knowledge base of ideas from the books I've read, where each idea is its own short note linked to the book it came from and to related ideas in other books.
Tech Stack:
Obsidian, Markdown, Quartz, GitHub Pages
Key Constraints:
Notes short enough to stand on their own, links that hold up across books, and publishing straight from my Obsidian vault.

I write my reading notes in Obsidian, one idea per note, and publish the vault as a static site with Quartz. Each book page lists its ideas under its backlinks, so a reader can start from a book and follow its ideas into the other books they connect to.

It covers eleven books so far, from Dale Carnegie's How to Win Friends and Influence People (1936) to Ryunosuke Koike's The Practice of Not Thinking (2010), including Blink, The Black Swan, Atomic Habits and Naked Statistics. The site has:

  • Graph view: a map of every idea and how it connects across books.
  • Library: every book and idea in one table.
  • Original ideas and definitions: my own ideas, and a glossary of the terms I've coined.