RAG · Agents · LLM Systems · Computer Vision
↑ Click the sphere
Who I Am
Building AI products that solve real problems — connecting model capability to user value through RAG, Agents, and thoughtful product design.
AI Product Building
0 to 1 product delivery
RAG Systems
Multi-stage retrieval pipelines
Agent Development
Autonomous AI workflows
LLM Integration
API-first architectures
Computer Vision
Detection & recognition
Full-Stack Dev
Next.js · FastAPI · SQL
Selected Work
An invite-only AI exam preparation platform for international UNSW students. Users upload course materials and the platform auto-generates summaries, quizzes, flashcards, and knowl…
Result: 175 tests passing; async AI generation eliminates gateway timeouts; 7 core feature modules covering the full study workflow, signi…
Led two AI products from 0 to 1 targeting fresh graduates, addressing fragmented job listings and low written/interview pass rates. Core KPIs: commercial GMV and DAU growth.
Result: Delivered a full-chain pipeline from job monitoring and auto-publishing to AI-assisted written exams and real-time interview coach…
A local voice AI assistant system designed for iOS/iPadOS, featuring a dual-agent architecture: Friday connects to a local private LLM via WebSocket for fully offline inference, wh…
Result: Delivered a fully offline voice AI pipeline — from wake word detection to local LLM inference to speech synthesis — installable as…
Engineering benchmark of four mainstream open-source Chinese TTS engines (CosyVoice2, Piper, MeloTTS, ChatTTS), measuring latency, naturalness, and feature support. Based on result…
Result: Built a complete TTS evaluation framework. The hybrid routing architecture reduces average response latency by 60%+, with a Flask…
An AI-powered intelligence dashboard for tracking tech influencers on X (formerly Twitter), built with twikit scraping and Claude AI analysis. Supports automated content collection…
Result: Delivered a full pipeline from automated KOL content scraping to AI analysis and visual dashboard, covering real-time tracking acr…
Built an agricultural pest recognition system on the AgroPest-12 dataset by benchmarking HOG+LR, YOLOv8, and Faster R-CNN. Optimized model architecture and training strategy for co…
Result: mAP@0.5 = 77.2%, outperforming the traditional HOG+LR baseline and supporting real-time field pest monitoring.
Built a global Chinese restaurant rating platform that quantifies the quality and distribution of Chinese cuisine in cities worldwide using a 100-point core index. Features include…
Result: Delivered a full-cycle platform from user submission to admin review, map visualization, and leaderboard ranking, with role-based…
Capabilities
Background
Focus on computer vision and machine learning.
Served as Head of Arts in the Student Union, responsible for event planning, team coordination, and cross-department collaboration.
Led multiple undergraduate projects, owning requirement breakdown, task allocation, and schedule management.
Built and applied cross-functional communication skills to effectively drive project delivery.
Get In Touch
◆Feel free to reach out through the channels above. Looking forward to collaborating in AI.