Alan Liang's Portfolio

About Me

I'm an AI/LLM software engineer and Computer Science graduate from UC Riverside, with specialized expertise in developing and evaluating AI-powered systems using cutting-edge frameworks. Currently pursuing my M.S. in Computer Science at USC, I combine deep technical knowledge with practical experience in fullstack development.

My work spans from LLM safety research—where I collaborated on advanced unlearning techniques to mitigate hallucinations and defend against adversarial attacks—to building production-ready applications. I'm passionate about creating tools that empower teams to launch fast and scale smart, with particular interests in Computer Vision applications in archaeology and LLMs for historical text analysis.

Technical Expertise

Python & PyTorch
LLM Development & Evaluation
React & Node.js
Machine Learning & NLP
Docker & Cloud Platforms
AI Safety & Optimization
Fullstack Development
Research & Analysis

Project Highlight

Verdant

AI-Powered Translation Platform & Verification System

In Development NPM Package Published

A comprehensive full-stack web application that combines AI-powered translation generation with intelligent quality verification for JSON localization files. Verdant features a dual-service architecture: an AI Translation Service that converts English JSON to multiple languages using OpenAI GPT-4, and a Translation Verification System that ensures accuracy using Google Gemini with RAG-enhanced context from professional translation datasets. Built on my custom NPM package with advanced vector search capabilities.

Tech Stack

React 19 Node.js TypeScript MongoDB Atlas OpenAI GPT-4 Google Gemini RAG + Vector Search Vercel + Render JWT Auth

Key Features

AI-powered JSON translation to 10+ languages
RAG-enhanced verification with professional datasets
Dual AI model architecture (GPT-4 + Gemini)
Real-time translation quality scoring
Vector similarity search for context
Automated error detection & correction
Batch processing & session management
Professional translation memory integration

🌐

Live Platform Try Verdant Online

📦

NPM Package @alnliang/tlv

-- Weekly NPM Downloads

1K+ Translation Pairs in RAG DB

-- Latest Version

Other Projects & Experiences

Broke Brokers

Stock Forecasting Platform

Jan. 2025 – Mar. 2025

Python React.js Node.js PyTorch

Collaborated in a team of 5 to design and develop a stock prediction dashboard, implementing an LSTM Machine Learning model with PyTorch for time-series forecasting.

Constructed a dynamic and responsive dashboard frontend using React and Vite, providing users with real-time stock data visualization.

Architected the backend infrastructure using Node.js and Flask, employing Express.js for RESTful API routing and Axios for seamless data fetching from external stock APIs.

Team PORT

Container Logistics Optimizer

Sept 2024 – Dec 2024

Python NumPy Pandas FastAPI

Led development on a container logistics optimization program within an Agile (Scrum) team of 5, utilizing Atlassian JIRA for task management, to determine optimal container loading/unloading sequences and ensure ship balance.

Single-handedly architected and implemented the core balancing and SIFT algorithms with sophisticated heuristics for the A* search algorithm, achieving a >50 percent reduction in search time for optimal logistical solutions—forming the backbone of the application's optimization engine.

Developed and exposed a RESTful API using FastAPI, enabling communication between the backend optimization engine and a web-based user interface (WebUI).

Professional Experience

Undergraduate Research Assistant

University of California, Riverside

Jan. 2024 – Jun. 2024 Riverside, CA

LLM Safety and Optimization Research

Investigated LLM unlearning techniques with PhD student under Dr. Salman Asif
Engineered evaluation programs using BLEURT model for semantic preservation analysis
Conducted literature reviews on state-of-the-art LLM safety methodologies

Alan Liang.