Ayaan Faisal Ayaan Faisal
portfolio

I like building things and making predictions;
Check out how I've applied my interests by developing tools aiming to tackle real problems ⬇

About Me

I’m apart of the Rutgers University Honors College, and for as long as I can remember, I’ve loved building things and seeing them function. Software engineering offered me a fast, creative, and non-resource-intensive way to do that.
I study Computer Science, Data Science, and Economics because I’m interested in how machine learning can solve real problems, and how those solutions connect to markets and decision-making. I’ve gained experience across full-stack development, data analysis, and finance-focused projects.

When I build, I start with problems I’ve seen firsthand and design with the people using the product in mind: users, customers, and teammates. I learn quickly, collaborate well, and love turning ideas into tools that actually get used.

Check out some of the work I've done and the story behind these products!

My Projects

Fantok

Fantasy Football, but with Stocks!

React, FastAPI, PostgreSQL, WebSocket, AsyncIO, Pandas, Alpaca

Fantok

Why Fantok?

We built Fantok because we noticed an increasing curiosity about financial markets and investing among high schoolers and early college students with no enticing avenue to explore the aforementioned ideas. With Fantok, we aimed to create a friendly and gamified way to explore the stock market and create strong portfolios. Check out its features below:

Features:

  • Real-time Trading: Integrates with the Alpaca Market API to fetch live stock data and execute simulated trades.
  • Fantasy League System: Users can create or join leagues, draft stocks in a real-time system, and compete against others.
  • Portfolio Tracking: Each user receives a virtual portfolio to track balances, holdings, and historical performance.
  • Weekly Matchups: Compete head-to-head with other members in your league based on portfolio growth.
  • Live Leaderboard: Dynamic leaderboard showing win rates, game stats, and overall rankings.
  • Modern UI: Built with React, TailwindCSS, and responsive design principles for a clean, mobile-friendly experience.

Skindex

AI-powered skin lesion classification using deep learning.

React, Flask, PyTorch, timm, ONNX Runtime, TensorFlow

Skindex

Why Skindex?

We built Skindex because my team and I could recount the unease associated with googling our symptoms, but not wanting to invest entirely in a visit to a medical clinic for mild skin concerns. So, we wanted a way to quell that unease, a way to get guidance about our skin conditions and whether or not we should commit to the full investment of a doctor's visit. Some features of the app include:

Features:

  • Ensemble AI Inference: Combines predictions from TensorFlow-based and PyTorch-based models, transported via ONNX to provide real-time classification with detailed confidence and agreement metrics.
  • 10-Class Detection: Identifies a wide range of common skin conditions, from critical issues like Melanoma to common ailments like Eczema and healthy skin.
  • Actionable Medical Insights: Maps results to specific urgency levels (High/Medium/Low) with tailored recommendations to guide user next steps.
  • Interactive Clinical Assistant: A smart chatbot that provides immediate diagnostic criteria, treatment protocols, and medical reference information based on user queries.
  • Modern React & Flask Architecture: Allows for image upload via file upload or camera, features a polished, responsive interface built with React 19 and TailwindCSS, powered by a high-performance Python backend.
  • Safety-Centric Design: Prioritizes user safety with clear medical disclaimers and guardrails, emphasizing that the tool is for informational support, not diagnosis.

Car Deal Predictor

Tackling Predictive Modeling With Kaggle Competitions

R, XGBoost, Random Forest, Data Cleaning, Ensemble Learning

Car Deal Predictor

Why Car Deal Predictor?

This project was a valuable personal pursuit, allowing me to explore machine learning principles and some of the fundamentals behind predictive modeling that I had initially only explored through classwork.

Features:

  • Data Cleaning: Handles missing values and standardizes categorical fields (Make, Model, Location).
  • Log Transformations: Applies log smoothing to skewed variables like Price and Mileage to improve model stability.
  • Ratio Metrics: Calculates engineered features such as price_per_km and price_to_fair_ratio to capture relative value.
  • Group Statistics: Computes aggregated mean price and value benchmarks grouped by Make, Model, and Location.
  • XGBoost Modeling: Implements Extreme Gradient Boosting with hyperparameter grid search and cross-validation.
  • Random Forest: Uses a Random Forest model to capture non-linear relationships as a complementary predictor.
  • Ensemble Strategy: Blends predictions from XGBoost and Random Forest using optimized weights (alphas) for maximum accuracy.

Contact Me