Compressed KV-Cache Layer for LLM Inference (Ongoing)
a compressed KV-cache layer for long-context LLM inference, integrating nano-vLLM with an LLM.265-inspired codec backend.
- Inference Engine
- KV-Cache Compression
All Projects
SyncNote - Collaborative Note Application
A collaborative Markdown editor featuring GitLab OAuth, sharing/co-editing, and live preview/syntax highlighting.
- Nuxt
- Vue
- MongoDB
- CI/CD
- K8s
LLM Fine-Tuning & Evaluation
An end-to-end fine-tuning & evaluation pipeline for gemma-3-270m, targeting abilities across factual QA, reasoning, and instruction following.
- PyTorch
- HuggingFace Transformers
- RAG
Frontdesk - Assistance Platform for Duke Hospital
An iOS App for medical trainees and faculty of Duke Hospital to facilitate information entry, communication, evaluation, and scheduling.
- SwiftUI
- Vapor
- PostgreSQL
Compiler Design and Implementation
A robust C compiler incorporating lexical analysis, syntax analysis, semantic analysis, IR generation, IR optimization, and code generation.
- C
- Lexer
- Parser
- IR generation/optimization
HiFOSS - Open-Source Forum Application
A cross-platform open-source forum App featuring post creation, comments, favorites, and private messaging.
- React Native
NBA Player Score Prediction
A predictive analysis applying different models such as MLP, Random Forest, and XGBoost to predict NBA player scores using player statistics from the 2023 season.
- Pandas
- scikit-learn
- XGBoost
Efficient Web Log Analysis with MapReduce}
An analysis program for large-scale website logs, extracting key website metrics and optimizing data processing efficiency.
- Hadoop MapReduce
- Java
RISC-V-based Computer System Development}
A RISC-V-based computer system on an FPGA board, incorporating instruction processing, keyboard interfacing, and VGA display modules.
- Verilog
- Quartus
- FPGA
- RISC-V