Dacheng Shen

Dacheng Shen

Computer Science Researcher

Passionate about AI, machine learning, and reinforcement learning, building intelligent systems that solve real-world problems.

🔬 View Research 💡 My Projects

🎓 Education

University of Southern California June 2024 - Dec 2025
M.S. in Computer Science — GPA 3.77/4.0
University of Connecticut Sept 2020 - May 2024
B.S. in Computer Science — GPA 3.65/4.0
Dean's List Honors

📄 Publications, Preprints & Presentations

DTRL: Reinforcement Learning Framework for Cyber-Physical Systems with Dynamic Time Requirements 2026
IEEE Real-Time Systems Symposium (RTSS), 2026 — Accepted
  • Yufeng Yang, Dacheng Shen, Mengyu Liu.
The Trafficker's Pitch: Detecting Deceptive Recruitment in Online Job Boards 2026
CySoc Workshop at AAAI ICWSM, 2026
  • Siyi Zhou, Peiran Qiu, Tanishq Salkar, Leonardo Blas Urrutia, Dacheng Shen, Nora Adadurova, Deyang Hsu, Eun Cheol Choi, and Emilio Ferrara.
A Simplex-Inspired Architecture for Integrating Quantum Capabilities into Cyber-Physical Systems Mar. 2026
Poster Presentation — 2nd HAIQ Workshop, Pittsburgh, PA
  • Tamim Ahmed, Dacheng Shen, Mengyu Liu, and Monowar Hasan. Poster presented at the 2nd HAIQ Workshop, Pittsburgh, PA, Mar. 2026.

🔬 Research

Washington State University

Research Assistant · Richland, WA · Jan. 2026 - Sep. 2026
Adversarial Robustness of RL-Based Quadruped Locomotion Apr. 2026 - Aug. 2026
  • Investigated observation-level adversarial attacks on terrain perception for RL-based quadruped locomotion, demonstrating physically induced failures through compromised dynamic balance.
  • Developed a reinforcement learning-based adversary that generates physically realizable, budget-constrained terrain perturbations, and improved robustness through adversarial retraining.
LLM-Guided Emergency Control for Power Grid Contingencies Jan. 2026 - Apr. 2026
  • Built an LLM + physics pipeline for IEEE 39-bus contingency mitigation, combining sparse generator selection, sensitivity-based SLP mapping, AC power-flow safety checks, and load-shedding fallback.
  • Benchmarked against AC-OPF and rule+SLP baselines across in-distribution and out-of-distribution N-1/N-2 scenarios with load perturbations.
  • Achieved 95% success while using substantially fewer control actions than AC-OPF with significantly lower evaluation cost.

University of Southern California

Research Assistant, FORTIS Lab (PI: Prof. Yue Zhao) · Los Angeles, CA · Mar. 2025 - Jan. 2026
PyHazards: AI-Powered Hazard Prediction Framework Sep. 2025 - Jan. 2026
  • Contributed Transformer-based models to PyHazards, an open-source Python library for multi-hazard prediction.
  • Developed modular data, training, and evaluation pipelines and benchmarked model performance against public baselines.
  • Standardized model interfaces and experiment workflows with the RAI Lab team to improve reproducibility and extensibility.
Comparative Narrative Analysis with Claude 3.7 Mar. 2025 - Jun. 2025
  • Developed an automated LLM-based framework using Claude 3.7 Sonnet for structured comparative narrative analysis under diverse prompting strategies.
  • Built scalable inference and response-processing pipelines through the Anthropic API, organizing outputs into structured JSON and human-readable text for systematic downstream analysis.
  • Implemented multi-dimensional evaluation across consistency, coherence, and bias to support systematic narrative analysis and human assessment.

University of Connecticut

Capstone Project Lead · Storrs, CT · Aug. 2023 - May 2024
OEM Controls Automated AS5 Angle Sensor Testing Aug. 2023 - May 2024
  • Led the development of an automated robotic calibration system for OEM Controls' AS5 angle sensor using Python, Ubuntu, and a UFACTORY xArm 6.
  • Integrated robotic-arm control, CANopen sensor communication, GUI-based operation, and automated result logging into a unified testing workflow covering a calibration range of −80° to 80°.

📚 Projects

Bullet-Hell Game — Software Design & Architecture Jan. 2026 - Apr. 2026
  • Developed a 2D bullet-hell shooter in Java, featuring multi-wave enemy spawning, a two-stage final boss, and Easy/Normal/Hard difficulty scaling.
  • Applied six OOP design patterns across 90+ classes — Strategy (IMovementStrategy, IBulletPattern with 8 behaviors), Factory & Builder (EnemyFactory, EnemyBuilder), State, and Command — within a layered codebase split into a reusable engine library and game project.
  • Implemented a data-driven wave sequencer, circle-based collision system, real-time HUD, and sprite animation manager, with service interfaces (IAssetProvider, ICollisionSystem, IEntityManager) for dependency injection.
Large-Scale Sentiment Analysis on Amazon Office Reviews Aug. 2025 - Oct. 2025
  • Cleaned and normalized 200K balanced Amazon review texts through HTML/URL removal, tokenization, stopword filtering, and lemmatization; trained unigram TF-IDF models with LinearSVC, Logistic Regression, and Perceptron, achieving 89.6% test accuracy.
  • Compared two GloVe representations using a feed-forward neural network—100-d average pooling and 1000-d concatenation of the first 10 tokens—showing that average pooling improved test accuracy from 77.7% to 83.3% and reduced overfitting.
Transfer Learning for Imbalanced Waste Classification Jan. 2025 - May 2025
  • Conducted a controlled ablation of three VGG16-based training configurations on the 9-class, 4,752-image RealWaste dataset, using stratified train/validation/test splits and a custom classification head.
  • Compared frozen-backbone and block5 fine-tuning strategies with inverse-frequency class weights and RandomOverSampler across three configurations; fine-tuning with cost-sensitive weights improved test accuracy from 60.6% to 69.9% and Macro F1 by 10.3 pp.
  • Performed confusion-matrix-based error analysis revealing persistent inter-class confusions, including Plastic/Metal and Paper/Cardboard, and proposed a class-selective hybrid imbalance strategy for future work.
AI-Driven Travel Assistant Design – FlySmart Mar. 2025 - May. 2025
  • Led a requirements engineering project to design an AI-powered flight booking assistant
  • Conducted stakeholder analysis, surveys, and interviews, derived user personas, empathy maps, and categorized requirements
  • Created and validated a Figma prototype with features including flight search, visa management, smart alerts, and personalized AI chat interface
  • Applied agile methodology with story-driven backlog, sprint-based development roadmap, and formal validation via usability testing and prototyping
Semantic Retrieval and QA System with Weaviate & RAG Sep. 2024 - Dec 2024
  • Built a semantic search pipeline using Weaviate with text2vec-transformers for vector-based brand similarity via GraphQL nearText queries.
  • Developed a PDF QA chatbot with Streamlit using PyPDFLoader, ChromaDB, and Hugging Face embeddings in a RAG framework, and benchmarked Weaviate vs. ChromaDB on retrieval accuracy and scalability.
  • Automated ingestion and querying with Python and Bash, orchestrating deployment with Docker Compose.
Student Admin Design Sep. 2023 - Dec. 2024
  • Collaborated on a group project, using Figma to design the product model and present the plan to stakeholders
  • Developed and implemented the code to bring each Figma design feature to life
  • Completed the website development and thoroughly tested all functionalities
Food Ordering App Design Aug. 2021 - Nov. 2021
  • Used Java to build a food ordering app that could generate an invoice with the name and price of the dish and the total price after ordering
  • Group leader, used GitHub to share and merge information and parts of what each group member had finished
  • Completed a report and made a presentation

💼 Experience

Software Development Engineer Jul 2023 - Aug 2023
HICOCA Intelligent Equipment Technology
  • Maintained web server backends and optimized database operations, ensuring data integrity, security, and system stability.
  • Developed and integrated new client-side UI features into the existing framework, significantly improving overall system functionality and user experience.

🔧 Skills

💻

Programming Languages

Python C# Java C/C++ SQL Bash/Shell HTML
🤖

AI/ML & NLP

PyTorch TensorFlow/Keras scikit-learn Hugging Face Transformers LLMs Transfer Learning RAG Semantic Search Prompt Engineering NLTK Gensim
🦾

Robotics & Simulation

Isaac Lab Isaac Sim NVIDIA Omniverse ROS 2 Vision-Language-Action (VLA)
🔬

Domain & Research

Cyber-Physical Systems Reinforcement Learning Adversarial Robustness Trustworthy AI Time-Series Forecasting
🗄️

Data & Databases

NumPy Pandas MySQL MongoDB ChromaDB Weaviate GraphQL Data Preprocessing
📐

Software Engineering

REST APIs OOP Design Patterns Modular Architecture Agile (Scrum) Git
🛠️

Tools & Platforms

Linux/Ubuntu Docker AWS Streamlit Jupyter Figma