About Me

Hi! I'm Tony Liu, a Computer Science student at Hunter College graduating in June 2026. I have a strong foundation in software engineering, IT Systems, and Machine Learning/AI, with experience building real-world applications using Python, React.js, Javascript, JavaScript, and cloud tools. Whether it's writing backend logic or managing infrastructure, I’m driven by solving problems that have impact. I’m open to roles in both Software Engineering and IT, especially where I can work hands-on with systems, build tools, and learn from a collaborative team. My goal is to keep sharpening my skills while contributing to real, impactful projects.

Projects

Spoken Vision: Multimodal Vision System for the Visually Impaired - Software Engineer

May 20, 2025

SpokenVision is an assistive AI application designed to help visually impaired users perceive their environment through real-time audio descriptions. It captures video input, detects and identifies objects, estimates their distance, and converts the information into spoken feedback using advanced AI models.

Spearheading development of a real-time vision system that provides spoken feedback to visually impaired users.

Contributed as a key member of a 3-person team to implement a multimodal AI pipeline integrating six pretrained models for object detection, depth estimation, image captioning, and scene-based Q&A.

Designed and developed an interactive React.js website with a backend API to display detection results and simulate end-to-end user interaction, enhancing accessibility and usability.

Uses camera feed to capture video, YOLOv8 for object detection, MiDaS for depth estimation, BLIP/Qwen for image captioning, spatial context modeling, and text-to-speech conversion.

Presented the application at the CUNY Tech Prep Graduation demo night event, where 100+ students and tech professionals tested and reviewed our project, providing positive feedback.

Food Chain - Software Engineering and Data Science Lead

December 11, 2024

A multipurpose application that combines a dish classification machine learning model, Retrieval Augmented Generation (rag) for recipe contexts, and AI-driven responses to provide comprehensive dish-related information.

Developed and fine-tuned a convolutional neural network using EfficientNetB0 as our base model for dish classification, achieving 74% accuracy across 101 dish categories.

Integrated a MongoDB vector database with over 25,000 recipes from Food.com and utilized LangChain for efficient retrieval of relevant recipe contexts.

Implemented a pipeline combining image classification, RAG, and OpenAI to generate accurate, context-aware dish and recipe suggestions.

Built a user-friendly interface with Streamlit and deployed the application on Huggingface for accessibility and performance.

Presented the application at a demo night event, where 100+ students and technical professionals provided positive feedback and reviews after testing the app.

FlashCardify - Full-Stack Developer

September 6, 2024

An innovative AI-based flashcard application designed to enhance learning through personalized flashcards.

Implemented AI-driven functionality to generate flashcards based on user prompts or uploads, leveraging machine learning to create personalized study materials.

Designed and integrated Create, Read, Update, and Delete (CRUD) operations for managing flashcards, decks, and user content.

Developed features for users to create, organize, and manage flashcard decks, allowing for structured and thematic study sessions.

Built an interactive quizzing system to test users' knowledge and track their progress over time.

Implemented sharing functionality to allow users to share decks with others, fostering collaborative learning.

Launched FlashCardify publically as an educational tool, using Next.js for fast performance, OpenAI for AI-driven flashcard creation, and Supabase for real-time database management.

Professor Finder - Full-Stack Developer

September 1, 2024

Developed a web application similar to RateMyProfessor, allowing users to search for professors based on parameters such as subject taught and star rating.

Implemented a chatbot feature that retrieves and displays detailed information about professors, including ratings, subjects, and user reviews.

Enabled class enrollment directly from the chat responses, allowing users to enroll in a professor's class with ease.

Built using Next.js with OpenAI API for our chatbot and Pinecone API for data storage and retrieval.

Retrieval Augmented Generation (RAG) Chatbot - Full-Stack Developer

August 27, 2024

Developed an advanced RAG-based chatbot for EarthSupremacy Corp, a fictional organization with a "take-over-the-world" theme

Designed and implemented the chatbot to retrieve relevant information from a database and generate coherent, context-specific responses in real time.

Deployed on Google Cloud Virtual Machine using the Next.js framework and uses Pinecone for data retrieval and management.

Travel Itinerary Generator - Software Engineer

February 7, 2024

A generative ai-based travel itinerary web application using the FERN (Firebase, Express.js, React.js, and Node.js) stack as part of Google mentorship program.

Utilized Google Gemini for itinerary generation, Pexels for image integration, JsPDF and HTML2canvas for itinerary saving, and Amadeus for flight lookup.

Deployed using Google Cloud Functions and Firebase and implemented user security through Firebase authentication and backend security using Express.js and Node.js middleware.

NYC Events Website - Software Engineering Lead

February 18, 2024

A real-time event tracker map developed during a HackCUNY hackathon using Firebase and React. This website allows users to create posts for events in NYC and view all events via Google Maps.

Integrated Google Maps API for location pinning, route planning, and real-time data visualization.

Implemented data management, user authentication, and user profile features using Firebase.

Led the project from planning to deployment, ensuring seamless user experience and backend functionality.