Hi, I'm William

Built "Dead Inside", a multiplayer VR game like Call of Duty: Zombies where players must survive against waves of zombies. Instead of using a gun to shoot the zombies, the zombies come up to you telling you about the problems they have in their lives and your task is to give them therapy.
Built the AI backend for character generation, conversations, speech-to-text, text-to-speech, and state management using Redis cloud.
The rest of the game is built in Unity with C#, including a p2p networking layer and Meta Quest 3 integration.
Ended up losing since the game was too ambitious and we didn't have enough time to finish it.

Built "AIFluence", a platform that creates virtual influencer personas, finds target audience, and creates content with complete life storylines and temporal consistency.
Automatically generates and schedules reels, stories, and posts on Instagram accounts to drive engagement through normal influencer content.
Enables businesses to advertise products through AI influencers or allows users to earn sponsorships with their virtual personas.
Won Best Startup 1st place ($7500)
Won Highest Startup Userbase Potential ($3500)

Made "Vibe Draw", an application that transforms simple sketches into 3D models.
Builtin integrations for drawing refinery, 3D model generation, and interactive 3D scene assembly.
Exports into a .gltf file for use in 3D modeling software.
Won Best Use of Gemini

👀 Those who know, know.
We were supposed to be trolling but somehow almost won something.

Created "Maple Valley" and analyzed a time-series and imbalanced dataset of venture capital investments, visualizing trends with plotly and streamlit and conducting predictive analysis with LSTM, ElasticNet, Random Forest, and Gradient Boosting Regressor and Classifiers. (Did not use ARIMA because of runQL)
Created "InsightOps", analyzing 18 million+ rows of user analytics data from amplitude with 30+ features using pyspark and pandas for data processing, streamlit and plotly for visualization, and used Markov Chains, Hidden Markov Models, ARIMA, LSTM, Prophet, and XGBoost for sequence prediction and time-series forecasting.
Won 4th place overall (6 finalists)
Won 6th place in runQL challenge
7th place in Federato challenge

Developed "Shallow Search", benchmarking the performance of Groq's Deepseek R1 70b distilled LLaMA model against GPT-4o in synthetic data generation using cosine similarity, specifically in the context of apache error logs.
Won Rootly Challenge

Built "CapitalQuest", a financial literacy game that teaches users about morgages, leasing, investing, stocks, and more. The game is built with react-three-fiber, drei, and react-three-rapier.
Won 1st place in SunLife case prize
Won top 5 overall

Built "Twirl" - Text to world, in real life. Twirl is an AI Agent with a web app interface that generates a 3D CAD model from a simple instruction with an optional image attachment. The app supports follow-up prompts for refinement and a separate parameter menu for tweaking the model. Used React.js, Next.js, three.js, tailwindCSS, drei, supabase, Azure blob storage, Azure image API, FastAPI, LangChain, LangGraph, and OpenSCAD + WebAssembly. The core feature of the app also includes sending a .stl file to a 3D printer and instantly printing the model.
Won 2nd place Overall
Won Best Use of GenAI

Developed a networking platform that stores facial embeddings as 128-dimensional tensors for recognition. The platform involves "battles" which are matches with 2+ people, where the goal is to network with the most amount of people at the event by scanning their faces and answering questions about them, in which the responses are validated by LLMs. Unfortunately, @arihan10 did not manage to finish the frontend using Unity because he was busy rizzing up AMD.

Developed "Haystack Diet", a web application that uses machine learning to plan events given data and geolocation. The app also provides personalized meal recommendations based on the provided dietary preferences and restrictions. It also integrates a game involving serving mustangs built with react-three-fiber that users can play to train themselves in understanding dietary restrictions
I could not compete with Ian Korovinsky and Fahmi Omer.

Made "Interview Monkey", a competitive platform for mock one-way interviews. The application makes interview prep more personalized, engaging, and interactive while simulating a competitive and realistic environment.

Created a fun, engaging, and rewarding way for users to reduce their carbon footprint and make eco-friendly decisions.
Won 1st place in RBC (Royal Bank of Canada) sponsor track

Developed an Adobe Express Add-on to empower content creators, marketers, and businesses to streamline their social media workflow without compromising on quality or creativity.
Won 3rd place Overall

Built a chrome extension that can instantly recognize text for sources of non-inclusive or discriminatory language and replace it with more welcoming vocabulary using Cohere's LLM.
Won Best Hack for Diversity & Inclusion
Won Honourable Mention: Use of AI
Won Honourable Mention: UI/UX

Built a mobile app with React Native that encourages users to achieve their goals from small deposits of AVAX (blockchain currency) as a sense of motivation.
My first hackathon, I did not know anything about React Native or blockchain.