Tiny Society

• A multi-agent AI social simulation and prediction engine, where AI agents observe, remember, reflect, and shift opinions over 7 to 30+ simulated days, then forecast where the world is heading.
• Built with Next.js, FastAPI (Server-Sent Events), a Python generative-agents engine, and optional Supabase saves with JWT auth.
• A consequence layer converts LLM social intents into calibrated affinity changes, giving relationships real inertia instead of model-written numbers.

WebsiteSource

LaunchPilot

• A two-time winning project at Hack Canada 2026 that moves a product from a brief or GitHub repo into a supervised, multi-agent launch workflow.
• Built with Next.js, FastAPI, SQLAlchemy, Alembic, and Postgres, with a Backboard memory layer giving each agent persistent context across runs.
• Maps competitors and pain points, generates ICP and messaging, builds a 7-day execution plan, and prepares outreach leads batched for approval.

WebsiteSourceDevpost

Markov Chain Model for Market Regime Forecasting

• A modular Python app that models daily equity return regimes (down/flat/up) with a first-order Markov chain and forecasts next-day state probabilities.
• Cleans price data, computes returns, discretizes regimes, and builds a row-normalized transition matrix for conditional next-state probabilities.
• Ships a CLI report tool and a Flask dashboard with threshold tuning, Monte Carlo simulation, and unit tests for the transition logic.

Source

Heat Mapping $5,000+ Thefts Around UofT (St. George)

• A full-stack crime intelligence app that maps Toronto Police theft-over-$5,000 incidents around the UofT St. George campus, built after my own laptop was stolen on campus.
• A Python/FastAPI pipeline filters police records by campus geospatial boundaries, normalizes the data in SQLite, and serves a clean API.
• A Next.js/TypeScript frontend renders an interactive OpenStreetMap heatmap with live summary metrics and incident sample tables.

WebsiteSource

Stochastic Risk Modeling: Gambler's Ruin Simulation

• A Gambler's Ruin simulation platform in Python that combines Monte Carlo experimentation with closed-form probability analysis to validate outcomes against theory.
• Organized as a reusable package (simulation, analytics, visualization, CLI, web) supporting up to 100k trials per run with Flask and Streamlit workflows.
• Includes convergence diagnostics, empirical-vs-theoretical error tracking, and an interactive Plotly dashboard for reproducible analysis.

Source