Yiming Cheng

EDUCATION

Tsinghua University, Department of Electronic Engineering Sep 2019 - Jul 2024

Bachelor of Engineering in Electronic Engineering (Major)

Minor in Statistics, Minor in Laws

University of Chicago, Department of Computer Science Expected Jan 2026

Master in Computer Science (Pre-doc),Mlsys track

GPA: 3.95/4.0

PUBLICATIONS

• Yang, Y., Cheng, Y., Feng, H., & Han, Z. "Emotion-Aware Scene Adaptation: A Bandwidth-Efficient Approach for Generating Animated Short". MDPI Sensors 2024..

• Lan, X., Cheng, Y., Sheng, L., Gao, C., & Li, Y. "Depression detection on social media with large language models". arXiv preprint arXiv:2403.10750.

• Lan, X., Piao, .., Cheng, Y., Gao, C., Li, Y., et al. "Recommendation for Inclusivity of Underrepresented Producers in User-generated Content Platforms". in Recycle.

• Yang, Y., Feng, H., Cheng, Y., Ma, Y., & Han, Z. "Minimizing Hallucinations and Communication Cost: Adversarial Debate and Voting Mechanisms in LLM-Based Multi-Agents". MDPI Applied Science 2025.

• Cheng, Y. "Research on Recommendation System Technology Based on Large Language Models". Tsinghua University Graduation Design.

• Patent: Yang, Y., Cheng, Y., Feng, H., et al. "A Semantic Encoding and Decoding Framework for Converting Visual Content into Virtual Animated Visual Representations".

RESEARCH EXPERIENCE

Graduate

Lmcache Team Sep 2024 - Present

Open Source Contributor Advisor: Prof. Junchen Jiang

LMCache: The first open-source Knowledge Delivery Network (KDN) that accelerates LLM applications up to 8x faster, at 8x lower cost
VLLM/production stack: Scale from single vLLM instance to distributed vLLM deployment without changing any application code. Now an official project under vLLM
Maintain and contribute to Open-Source repo. Working on Router Scheduling and cloud deployment. Contributed 1262 lines of code

Argonne National Laboratory May 2025 - Oct 2025

Research Assistant Advisor: Prof. Kexin Pei

2025 Summer of Reproducibility (SoR) Fellowship under the Open Source Research Experience (OSRE) and REPETO programs,with United States National Science Foundation
EnvGym: Project Leader,build and design a system-optimized agent system for environment setup

Undergraduate

Future Intelligent Lab (FIBLAB), Tsinghua University Jul 2022 - Jun 2024

Research Assistant Advisor: Prof. Yong Li

Recommendation for Inclusivity of Underrepresented Producers in User-generated Content Platform:

Take the pioneering step to thinking of the inclusivity issue of underrepresented producers in UGC (user-generated content) platform
Propose to construct a heterogeneous graph that can enrich the relations of vulnerable populations, and further propose graph neural networks to learn representations based on enriching features from multi-hop neighbors

City Socioeconomic Simulator based on Large Language Models:

Use UE to Build a visual model scene of Beijing (CBD district)
Use python to write scripts for agents to interface with LLM and design the agents' memory mechanism to do POI recommendation (POI means point of interest in the city)
Design and plug in agent-based recommendation systems

Signal Processing Lab, Tsinghua University Mar 2022 - Jun 2024

Research Assistant Advisor: Aso Prof. Yi Yang

Emotion-Aware Scene Adaptation: A Bandwidth-Efficient Approach for Generating Animated Shorts

Use the PyTorch framework, build an image element and emotion recognition model based on the CLIP model and InceptionV3, and use PAD (Pleasure-Arousal-Dominance) for emotion scoring
Enhance the generated semantics using the EmoCap model trained based on PAD scores for emotion style, ultimately achieving higher emotional coherence than the baseline on the received new video frames

Wireless Networking, Signal Processing and Security Lab, University of Houston Apr 2022 - Jun 2024

Research Assistant Advisor: Prof. Zhu Han, Member of NAS

Scalable AI Generative Content for Vehicular Network Semantic Communication

This project aims to establish a large-model-based semantic communication channel and test its accuracy on a vehicular dataset
Build and test a channel in PyTorch that uses CLIP to convert original images into semantics and then uses Stable Diffusion to restore semantics back into images

INTERNSHIP EXPERIENCE

Beijing SmartBow Information Technology Co., Ltd. Jun 2023 - Sep 2023

Software Engineer

Refactor the Sunflower library (main functions include JSON parsing, MQTT, B-Stack device information parsing, and data transmission encryption) for the company's Internet of Things (IoT) data platform using Go-lang
Perform functional and performance testing on the refactored Sunflower library
Collaborate with hardware interns to debug and ensure successful MQTT-based data transfer of bridge deflection, vibration frequency, and temperature data from LuZhou Bridge to the company's database

Beijing Thunisoft Information Technology Co., Ltd. Jul 2022 - Sep 2022

Software Engineer

Use Spring Batch to develop a batch job scheduling system supporting complex workflows and dependency management. Scheduled tasks are executed as planned using Cron expression triggers
Integrate Quartz scheduler for enhanced flexibility
Data integrity and stability are assured with Spring transaction management and JDBC operations

ADDITIONAL INFORMATION

Scholarship: Merit-based Predoc Scholarship of $40,000, University of Chicago (2024)
United States National Science Foundation for SoR project (2025)

Research Fields:

Undergraduate: Data mining (Recommendation System, Emotion Awareness, EmbodiedCity)

Current/Future: System for machine learning (distributed LLM deployment, distributed KV cache, efficient ml), Machine learning for systems (machine learning for code generation)

Programming Skills: Python (PyTorch, CuPy), Go (Docker, K8s), Git (Github action), Linux, C, C++, Matlab, Verilog etc.

Personal Website: https://eaminc.github.io/ (includes GitHub, Google Scholar and other detailed information)

EDUCATION

PUBLICATIONS

RESEARCH EXPERIENCE

INTERNSHIP EXPERIENCE

ADDITIONAL INFORMATION