You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: src/data/personalInfo.ts
+21-11Lines changed: 21 additions & 11 deletions
Original file line number
Diff line number
Diff line change
@@ -24,33 +24,43 @@ export const personalInfo = {
24
24
],
25
25
experience: [
26
26
{
27
-
title: "AI Engineer Intern (Incoming)",
27
+
title: "AI Engineer Intern",
28
28
company: "DIET, Harvard University",
29
-
duration: "Sep 2025 – Dec 2025",
30
-
description: "Developing an end-to-end AI pipeline to assist in the identification of stolen ancient coins using image segmentation, embedding-based retrieval, and auction data integration."
29
+
duration: "Aug 2025 – Present",
30
+
description: "Architected an agentic system and fine-tuned multimodal VLMs for artwork analysis at Harvard Arts Museum.",
31
31
},
32
32
{
33
33
title: "Machine Learning Engineer Intern",
34
34
company: "Institute of Artificial Intelligence, Nanjing Normal University",
35
-
duration: "Aug 2023 - Jun 2024",
36
-
description: "Independently developed a diffraction-pattern recognition algorithm using computer vision and deep learning to automate RHEED image analysis, leading the end-to-end algorithm design and implementation."
35
+
duration: "Aug 2023 – Jun 2024",
36
+
description: "Designed a production-grade computer vision system for RHEED image analysis; led a 6-member team and achieved substantial time and cost savings.",
37
37
}
38
38
],
39
39
projects: [
40
40
{
41
-
name: "AI and Education",
42
-
description: "Led the team to design modular AI solutions for educational applications, focusing on model architecture and content generation pipelines. Published paper 'A Comprehensive Investigation for ChatGPT's Applications in Education'.",
description: "Agentic system that orchestrates 20+ V&L models/APIs (Qwen-VL, BLIP-2, GPT-4o, Gemini, YOLOv8) to analyze user-uploaded artwork images, producing structured metadata; reduced cost by 40% and improved speed by 58%.",
name: "Multimodal VLM Fine-Tuning for Artwork Description",
47
+
description: "Fine-tuned Qwen2.5-VL-7B with LoRA on curated artwork datasets; leveraged ViT and Qwen-VL on 385K+ images to generate 70M+ tags/descriptions using warm-up, cosine LR decay, BF16, and gradient accumulation.",
48
+
technologies: ["Qwen2.5-VL-7B","LoRA","SFT","BF16","Cosine LR Decay","Warm-up","Gradient Accumulation","ViT"]
49
+
},
50
+
{
51
+
name: "AI Chat Website with RAG System",
52
+
description: "Developed a full-stack AI chat application using React, TypeScript, and OpenAI GPT API. Implemented RAG (Retrieval-Augmented Generation) with ChromaDB vector database, MMR algorithm for diverse retrieval, and deployed on Railway backend with GitHub Pages frontend.",
description: "Performed data cleaning, exploratory data analysis and visualization on BPD budget and payroll records to uncover trends in overtime spending. Built and evaluated machine learning models to predict future overtime expenditures.",
description: "Built an accessibility assistance system using YOLO-based object detection and semantic segmentation for blind pathways. Responsible for technical architecture, dataset collection, and model development.",
description: "Built an accessibility assistance system using YOLO-based object detection and semantic segmentation for blind pathways. Responsible for technical architecture, dataset collection, and model development.",
0 commit comments