Back to exploring

THE 30-SECOND VERSION

Donald Shen.

Computer Engineering · University of Illinois Urbana-Champaign

Donald’s work spans autonomous driving, embedded driver monitoring, machine learning, local LLM inference, industrial SOP verification, software systems, quantitative finance, and interactive learning tools. Start with the leading project in each field, then use a role lens to narrow the evidence.

A lens, if you like

The lens changes the order of the work. The evidence stays the same.

STRONGEST RELEVANT EVIDENCE

Robotics & Autonomous Driving · FIELD RANK 1

Autonomous driving lab

Robotics & Autonomous Driving · Project record · Sep 2026

Donald implemented lane-segmentation stages, curvature-based speed control, particle-filter updates, eight-direction LiDAR support, and an offline localization parameter sweep within shared course infrastructure.

An integrated set of perception, control, and localization modules, with training checkpoints and simulation experiments that compare position error, heading error, convergence, and effective sample size.

Embedded Systems & Computer Vision · FIELD RANK 1

Driver monitoring & diagnostic lab

Embedded Systems & Computer Vision · 2025 – Present

Deployed and validated real-time inference for eight safety-relevant detector classes. Designed the hierarchical alert state machine and vehicle-context gating, persisted per-camera calibration profiles, and built the diagnostic lab CLI.

Validated on real cabin video and shipped a 616-test pytest suite with replay-driven regressions, inference-cadence checks, and per-detector coverage metrics in lockstep with model development.

Machine Learning & Model Research · FIELD RANK 1

Time-series forecasting & retrieval

Machine Learning & Model Research · February–March 2026

Donald built raw-feature retrieval baselines, model-adaptation experiments, causal PCA, and persistence-head training around upstream time-series models. The related Kronos RAG workspace is a separate local study.

The research framework connects retrieval, forecasting, and decision-rule evaluation. Ablations exposed an objective that disabled simulated exits, while regime-selected retraining did not outperform its rolling baseline in the documented study.

Local LLM Inference & Agents · FIELD RANK 1

Local Qwen inference

Local LLM Inference & Agents · Project record · Sep 2026

Application-side controls wrap upstream Qwen/vLLM with durable request budgets, fixed decoding settings, runtime metadata, and structured outputs. Donald also contributed private-serving configuration in the event-ending project.

Local Qwen generation was exercised successfully. The client handles output truncation and distinguishes transport, HTTP, parsing, and completion failures; this is inference integration, not a new model or serving engine.

Software Systems & Automation · FIELD RANK 1

Illinix 391

Software Systems & Automation · Spring 2025

Architected a kernel from scratch with virtual memory, fork/exec, 15+ system calls, VIRTIO block drivers, and a custom KTFS filesystem.

A complete kernel supporting process creation, executable loading, filesystem access, multitasking, and an interactive shell.

Quantitative Finance & Markets · FIELD RANK 1

Prediction-market execution systems

Quantitative Finance & Markets · September 2026

Donald’s agent-assisted independent refactor separates execution, tape, backtesting, research, and dashboards, with explicit runtime paths and write controls. Earlier system and research snapshots retain shared-source attribution.

An independently packaged execution/research system with reconciliation for ambiguous order outcomes. Dated September 4 verification records 1,526 passing offline tests; production equivalence and profitability are separate, unproven outcomes.

Interactive Tools, Learning & Writing · FIELD RANK 1

twentyfourpoints.com

Interactive Tools, Learning & Writing · Summer 2025

Built the React/TypeScript client, Node.js and Socket.io multiplayer backend, Supabase authentication and ELO rankings, and a custom mathematical game engine.

Launched in 72 hours. Donald’s Summer 2025 project record reports 760+ unique puzzles solved and 100% uptime during the reported launch period.

PROFESSIONAL & RESEARCH EXPERIENCE

2025 – PresentRemote

Shanghai Ben’an Intelligent

Research Engineer, Driver State Monitoring (DMS) · Internship

Deployed and validated real-time on-device driver-state monitoring, connecting safety-relevant alerts, low-power edge inference, and replay-driven validation.

  • Deployed and validated eight safety-relevant detector classes: eye closure, yawning, distraction, phone use, smoking, face loss, lens occlusion, and eye anomaly. Shipped a 616-test pytest suite with replay-driven regressions, inference-cadence checks, and per-detector coverage metrics alongside model development.
  • Designed the alert state machine with hierarchical face-loss and lens-occlusion fallbacks. Speed, ignition, and gear gating suppress false alarms in non-driving states; persisted per-camera calibration profiles support fleet-grade traceability.
  • Fit the model pipeline to the low-power CV181x edge SoC by tightening inference cadence and per-stage compute budgets. Validated end-to-end on real cabin video and built the lab CLI for shadow-mode hard-case capture, human review, and replay-based metric reporting.
On-device inferenceCV181xPythonpytestState machinesCalibrationReplay validation
2021RESEARCH

Shanghai Jiao Tong University

Autonomous Driving Research

Implemented YOLOv3 object detection, OpenCV and SLAM, and lane-following algorithms on a ROS platform for autonomous navigation.

YOLOv3OpenCVSLAMROSAutonomous navigation
2021Remote

McKinsey & Company

Part-Time Research Analyst

Produced McKinsey-standard research reports on industrial automation and inventory management systems.

Industrial automationInventory managementResearch synthesis

PUBLICATIONS

ACADEMIC SPECIALTIES

Summer 2022EDUCATION

University of California, Berkeley

Summer Session, EECS

Summer study in electrical engineering and computer sciences.

Currently studying · Fall 2026
ECE 385 Digital Systems Laboratory · ECE 408 Applied Parallel Programming · ECE 490 Introduction to Optimization · ECE 498 AI Systems & Engineering · IE 421 High Frequency Trading Technology

Hardware & circuits

ECE 110 Introduction to Electronics · ECE 120 Introduction to Computing · ECE 210 Analog Signal Processing · ECE 330 Power Circuits & Electromechanics · ECE 342 Electronic Circuits

Systems & algorithms

CS 225 Data Structures · ECE 220 Computer Systems & Programming · ECE 374 Introduction to Algorithms & Models of Computation · ECE 391 Computer Systems Engineering

Robotics & control

ECE 470 Introduction to Robotics · ECE 484 Principles of Safe Autonomy · ECE 486 Control Systems

Machine learning

ECE 364 Programming Methods for Machine Learning · ECE 449 Machine Learning · ECE 498 Deep Generative Models · CS 441 Applied Machine Learning

Mathematics & optimization

ECE 313 Probability with Engineering Applications · MATH 213 Basic Discrete Mathematics · MATH 241 Calculus III · MATH 257 Linear Algebra with Computational Applications · MATH 285 Introduction to Differential Equations

Physical sciences

PHYS 211 University Physics: Mechanics · PHYS 213 University Physics: Thermal Physics · PHYS 214 University Physics: Quantum Physics

Economics

ECON 103 Macroeconomic Principles (Exam credit) · IE 421 High Frequency Trading Technology (In progress · Fall 2026)

Full coursework & current study

START A CONVERSATION

Let’s connect the work to your team.

For a current résumé, availability, or a conversation about a specific role, contact Donald ↗.

SUGGESTED INTERVIEW QUESTIONS

  1. How did virtual memory, fork/exec, and preemptive scheduling fit together in Illinix?
  2. How did you keep DMS camera capture and recorded replay on the same timing and detector contracts?
  3. How do lane perception, curvature-aware speed control, and particle-filter localization interact?
  4. How does the SOP verifier distinguish a legitimate alternate path from a skipped prerequisite?
  5. How does the C++ runtime avoid resuming a task before its context is safely parked?
  6. What belongs in a local-inference budget, and what do you do with invalid model output?
  7. What did a forecasting objective optimize successfully while breaking the downstream exit rule?

Updated September 2026. Project sources, contribution context, and reported outcomes are linked from each record. Experience supplied by Donald; publication metadata links to IEEE records.