Timothy Lee.
Aspiring ML Engineer.

Visit Project

AI-powered code navigation tool helping developers understand codebases through natural language queries. Features semantic code search, interactive graph visualisation of file dependencies, RAG-based repository indexing, and contextual AI chat. Built with Next.js, OpenAI, React Flow, and Upstash for vector search and caching.

Developed a comprehensive security platform featuring real-time object detection using YOLOv11, Llama 3.2, Florence, and SAM2 models. Implemented an automated training pipeline for CV models and built an interactive admin dashboard with live patrol mapping and incident tracking. Leveraged DeepSort for object reidentification to support efficient policing operations.

Developed a URL shortener using Golang and PostgreSQL, implementing a RESTful API for efficient URL management. Enhanced user experience with link expiry and fast redirects with Redis cache. Deployed on Vercel for scalability and reliability.

Competed in a machine learning competition, achieving a 94% anomaly detection AUC-ROC score by employing deep neural networks, logistic regression, and one-class SVM. Enhanced model performance through extensive feature engineering, utilising techniques such as SMOTE and ELU, successfully identifying and mitigating anomalies in a recommendation system.

Developed and deployed 'SG read already?', a web application enhancing accessibility with document summarisation and scan-to-speech for the visually impaired, using Next.js. Led the creation of an accessible platform for enhanced document understanding, catering to elderlies and individuals with visual challenges, demonstrating commitment to digital inclusivity.

Designed a conversational AI Streamlit Python app using Langchain and Hugging Face models. Enhanced user experience by incorporating features like Conversational Memory, dynamic interaction, and session management, improving context awareness and user experience. Utilised Retrieval Augmented Generation for precise and relevant real-time FAQ responses.

Designed, developed, and deployed a comprehensive university guide using the MERN stack. Integrated modern UX techniques, responsive design, ensuring an engaging user experience. Seamlessly incorporated OAuth, email notifications, and OpenAI's GPT API.

Integrated SERP and OpenAI's GPT API in Python to automate article curation and summarisation for optimised Instagram content, employing Langchain for strategic information distillation. Enhanced social media engagement by transforming AI-condensed articles into compelling Instagram posts with targeted hashtags for amplified reach.

Developed an interactive Streamlit Python app designed to enhance users' image journey by offering advanced features such as object detection, image captioning, and metadata tagging.

Harnessed the power of natural language processing using Keras, machine learning and deep learning to scrutinise news articles' credibility, yielding 95% accuracy rate on a dedicated dataset.