DISTRIBUTED SYSTEMS RESEARCHER / Software Engineer

Hi, I am Jieru Wang.

I design intelligent, high-concurrency systems by combining deep reinforcement learning with edge intelligence—focusing on dynamic resource allocation, cloud-edge synergy, and system resilience."

Explore my work Download resume

Edge Intelligence System Design Chaos & Fault Tolerance

About

Curiosity, compassion, and computation guide everything I build.

Professional Focus

With a Master's degree from Ocean University of China (GPA 92.2) and enterprise engineering experience at DiDi Global (Beijing, China), I specialize in architecting intelligent, high-availability distributed systems. My expertise combines DRL-driven edge resource optimization in academic research with cloud-native chaos engineering in production—focusing on dynamic scheduling, automated fault injection, and system resilience.

Go & Python engineering
Edge AI & DRL Optimizations
AI-Augmented Engineering (MCP/Agents)
System Resilience & Reliability

Engineering & Research Philosophy

During my undergraduate years, I honed my coding and algorithmic skills and participated in numerous programming competitions, which greatly inspired my future research and career direction. I felt the charm of algorithms and gradually learned to strike a balance in algorithm optimization and computer networks through wonderful thinking and integration. Since joining Didi for over a year, I worked with the team to improve the fault injection pipeline. I understand that in complex systems, any small error can have a fatal impact, especially in a real-time service like Didi, where the reliability of the system must be improved.

"True system intelligence lies not just in optimal decision-making, but in remaining resilient when things fail."

Work & Projects

From research prototypes to scalable platforms, here is what I am building.

Illustration: a fault is injected into a Kubernetes cluster of microservices, and the platform verifies automatic recovery

DiDi Global | Enterprise Cloud-Native Chaos Engineering System

Built and upgraded DiDi’s enterprise-scale chaos engineering platform to evaluate microservice resilience and automated fault recovery under high-concurrency production workloads.

Cloud-Native Event-Driven Architecture: Re-architected the execution pipeline using Kubernetes CRDs, Go Operators, and gRPC Daemons for scalable, asynchronous fault injection.
Kernel-Level Resilience & Fault Control: Managed application, middleware, and Linux kernel-level faults (e.g., tc/netem network delay, packet loss, and cgroup stress).
Automated State Consistency: Built transactional locking and side-channel verification mechanisms to guarantee zero orphaned fault rules during system crashes.

Go / Kubernetes CRDs
gRPC & Linux Kernel (tc)
System Reliability & Consistency

Illustration: risk signals are detected, analysed by an AI copilot for root cause, and resolved automatically, driving incident counts down

DiDi Global | Intelligent Risk Governance & Incident Platform

Engineered enterprise-scale risk management and incident governance platforms to manage system vulnerabilities, RCA, and fault recovery across company-wide microservices.

Full-Lifecycle Incident & Risk Governance: Built automated risk pipeline via Go, event-driven DDMQ, and Spark, covering risk detection, ticket generation, and SLA metric tracking.
MCP-Driven AI Incident Copilot: Leveraged Model Context Protocol (MCP) and LLM agentic workflows to automate fault investigation paths, root-cause analysis (RCA), and incident post-mortem generation.
Proactive Threat Mitigation: Driven the strategic shift from manual incident handling to automated risk discovery, significantly reducing online outages and SLA consumption.

Python / Spark
Go / MQ
AI-Augmented Engineering (MCP/Agents)

Illustration: stacked CT slices are processed in parallel into an interactive 3D model in the browser for surgical puncture simulation

3D Surgical Simulation & Medical Imaging Platform

Engineered a high-performance smart healthcare system for 3D surgical simulations and automated medical imaging workflows.

Asynchronous Model Processing: Designed a multi-threaded task pool to call AI model algorithms asynchronously, resolving critical performance bottlenecks in synchronous execution.
Interactive 3D Visualization: Built a web-based 3D surgical rendering pipeline using Vue 3 and VTK.js, enabling real-time CT scene visualization and surgical puncture simulations.
Secure Session Architecture: Implemented JWT authentication and Redis-backed session management for persistent doctor access and reliable CT model metadata storage.

Spring Boot / Redis / MySQL
Docker orchestration
Vue 3 / VTK.js visualization

Research Highlights

Bridging deep reinforcement learning algorithms with edge intelligence and dynamic resource allocation.

COMPUTER NETWORKS (2025)

DRL-Based Task Offloading in Wireless VR

Proposed the Task Prediction and Multi-objective Optimization Algorithm (TPMOA) to minimize user waiting time, rendering latency, and energy consumption in Wireless Virtual Reality (WVR) edge networks.

Viewpoint Prediction: Integrated a local prediction model into WVR systems to reduce computational load while maintaining high accuracy in high-dimensional spaces.
DRL Offloading Engine: Combined entropy mechanisms and representation learning to optimize real-time edge offloading, reducing latency by 11.39% and energy by 3.99%.
Research Output: First-authored paper published in Computer Networks (CCF-B, JCR Q1) and granted 1 national invention patent.

COMPUTER COMMUNICATIONS (2025)

Multi-Agent DDPG for Federated IoT Allocation

Designed a multi-agent reinforcement learning framework to optimize long-term resource allocation for Federated Learning (FL) across heterogeneous IoT devices.

Heterogeneity-Aware Modeling: Built a resource allocation architecture addressing dynamic bandwidth variations, computational limits, and heterogeneous IoT hardware.
MAEDDPG Algorithm: Developed an enhanced Multi-Agent DDPG method combining LSTM, noisy networks, and double critic networks to accelerate FL task convergence.
Research Output: Co-authored paper published in Computer Communications (CCF-C, JCR Q2)

Life in Motion

Balance fuels creativity. These are the spaces that keep me grounded.

Pilates Practice

I enjoy the slow and powerful relaxation that Pilates brings to me.

Tennis

Tennis is a new sport that I started trying out last year. Every time I hit the ball, I feel extremely excited.

Skating

Skating is the most elegant and free sport, because there is almost no resistance on your feet.

I Ching

This is a very ancient fortune-telling system in China. When you are confused, you might as well ask the ancients for the answers.