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
summary: "A full-stack platform that helps connect NYC residents with vocational programmes: Django-powered backend (PostgreSQL, WebSockets), Google Maps API for geo-search, AWS Elastic Beanstalk + Nginx deployment, and a Travis-driven CI/CD pipeline."
- name: Information Retrieval System & Retrieval-Augmented Generation
35
+
timeline: "Sep. 2024 – Dec. 2024"
36
+
summary: Built a search engine from scratch on MS MARCO—block-compressed inverted index, DAAT BM25 ranking—then refactored the pipeline with HNSW and advanced reordering to lift F1 by 25 % and systematically probe “lost-in-the-middle” bias in RAG workflows.
- name: Semantic Segmentation by Pixel-level Time Series Classification
27
40
# logo: /images/sections/projects/STC.png
28
41
# role: First author
@@ -46,28 +59,28 @@ projects:
46
59
timeline: "Sep. 2023 - Nov. 2023"
47
60
repo: https://github.com/yang-i-hu/TS_US_Candy/
48
61
summary: The project focuses on forecasting the candy production in the U.S. based on the time series data over the past 45 years. Techniques such as Box-Cox transformation and differencing are applied to achieve stationarity, and the optimal SARIMA model is identified using ACF/PACF analysis and Maximum Likelihood Estimation. The model are further validated through comprehensive diagnostic tests and spectral analysis.
summary: Developed an automated workflow to scrape and process PDFs from a government website. Employed template matching and a custom-trained TrOCR model to extract handwritten data
- name: Efficient Visual Attention Design for Image Super-Resolution
57
65
logo: /images/sections/projects/evasr.png
58
66
role: AAAI 2024 submission
59
67
timeline: "Mar. 2022 - May 2023"
60
68
repo: https://github.com/yang-i-hu/EvaSR
61
-
summary: Replicated and analyzed 16 Super-Resolution models to assess key characteristics of success models. Collaboratively designed a CNN-based model that achieved state-of-the-art performance while reducing parameter count by 85% through the innovative use of efficient visual attention mechanisms.
62
-
tags: [ "Deep learning", "Computer Vision", "Image Super Resolution", "Pytorch","Python"]
69
+
summary: Replicated 16 SOTA SR models, discovered key architectural bottlenecks, and co-created EvaSR—an attention-augmented CNN that matches SOTA PSNR while slashing parameters by 85 % and FLOPs by 70 %.
summary: Automated the collection and parsing of thousands of government PDFs, then fine-tuned a TrOCR-base model with custom template-matching to hit 95 %+ character accuracy on challenging handwritten forms—reducing manual data entry time from hours to seconds.
- name: Soccer player transfer market value prediction
65
78
logo: /images/sections/projects/evasr.png
66
79
role: Machine Learning Course Project
67
80
timeline: "Sep. 2022 - Jan. 2023"
68
81
repo: https://github.com/yang-i-hu/TMP
69
-
summary: The project focuses on predicting the market value of soccer players from the top 5 leagues in the transfer market based on their match performance statistics. Eight machine learning models were tuned, including KNN, Random Forest, and Gradient-boosted Trees. The results of these models are reported along with an Exploratory Data Analysis using R markdown.
70
-
tags: [ "Machine Learning", "R"]
82
+
summary: Engineered a rich feature set from top-5-league match stats and benchmarked eight machine learning models including KNN, Random Forest, and Gradient-boosted Trees. The results are reported along with an Exploratory Data Analysis using R markdown.
83
+
tags: [ "Data Science", "Machine Learning", "R"]
71
84
72
85
- name: Exploring SO(3) through the lens of Orbifolds
summary: A tiny script that may help deal with the tedious calculations in differential geometry.
94
+
summary: Authored a lightweight Python tool that symbolically computes curvature, torsion, and Christoffel symbols—turning pages of manual tensor algebra into one-line scripts.
0 commit comments