Welcome to my project portfolio!
I'm a Computer Science student and Apple Training Lead passionate about building intelligent systems that connect humans and machines through clear, efficient, and empathetic design.

Built at HackPrinceton 2025
Repository Link

Description:
Recall is an intelligent facial recognition system designed for profile management and timeline integration. It leverages computer vision and machine learning to detect, store, and manage facial data with seamless API integration for IoT devices.

Tech Stack:
Python • Flask • MongoDB • OpenCV • Raspberry Pi Integration

Key Features:

• 📸 Face detection and profile recognition using OpenCV
• 💾 MongoDB schema for secure profile and timeline storage
• 🔌 Flask API for CRUD operations on facial profiles
• 📤 Photo upload and management capabilities
• 🤖 Raspberry Pi integration endpoints for real-time facial recognition

My Role:
Built the facial recognition API backend, implemented MongoDB schema design, and created Flask endpoints for profile management with photo upload functionality.

Built at HackHarvard 2025
Devpost Link

Description:
Offscript bridges the gap between technical knowledge and communication by simulating real coding interviews with AI-driven feedback. Users speak through problems while Offscript evaluates clarity and problem-solving approach.

Tech Stack:
Next.js • FastAPI • Gemini AI • Vapi • TailwindCSS • SQLite

Key Features:

• 🎤 Voice-based technical interviews with real-time feedback
• 🧩 AI analysis of communication clarity and problem-solving flow
• 📈 Performance analytics with transcript playback
• ⚙️ Invisible metadata streaming for seamless code context tracking

My Role:
Led backend AI orchestration, integrating Gemini AI to analyze user explanations and streamline code-context synchronization.

Tech Stack: Python • TensorFlow • Librosa

Highlights:

• Built end-to-end audio preprocessing and model training pipelines.
• Integrated agentic-system logic for real-time emotional state prediction.
• Achieved a 30% gain in inference efficiency through ETL optimization.

📂 Repo: github.com/andreay99/sona-ai

Tech Stack: TypeScript • React • Node.js • OpenAI API

Highlights:

• Built a backend ML module using PyTorch to detect and fix runtime bugs.
• Designed an interactive web interface that explains each fix step-by-step.
• Achieved a 50% improvement in error resolution speed.

📂 Repo: github.com/andreay99/gpt-code-debugger

Tech Stack: Python • CodeT5+ • Hugging Face Transformers

Highlights:

• Curated domain-specific datasets for supervised fine-tuning.
• Improved model accuracy by 25% using custom prompt templates.
• Developed scalable pipelines for text-classification and code-review feedback.

📂 Repo: github.com/andreay99/code-feedback-llm

Tech Stack: SQL • PostgreSQL

Highlights:

• Designed a relational database for large-scale trip analysis.
• Implemented queries to detect anomalies and identify >30 min ride patterns.
• Delivered insights for improving city-wide bikeshare operations.

📂 Repo: github.com/andreay99/bikeshare-trip-analysis

Languages: Python • JavaScript / TypeScript • Java • SQL • HTML / CSS
Frameworks: React • Node.js • FastAPI • TensorFlow • PyTorch • LangGraph
Databases: PostgreSQL • MySQL • SQLite
Concepts: NLP • Agentic Systems • ML Model Lifecycle • Data Modeling • Prompt Engineering

• 🎓 B.S. Computer Science — NJIT (2025 – 2027)
• 💼 Training Lead @ Apple (Edison, NJ)
• 🔍 Passionate about building accessible, human-centric AI systems

Let's Connect:
🌐 andreay99 on GitHub
💼 LinkedIn
📧 andreayanez11@outlook.com

All projects in this repository are released under the MIT License unless stated otherwise.