ML Engineer & AI Researcher

Dianmu
Zhang Ph.D

Agentic AI · Computer Vision · Edge ML · Robotics

Building and shipping AI systems that reach billions of users — from real-time inking prediction on Windows to multi-agent LLM systems for healthcare. PhD from University of Washington. 5 granted & pending patents. Published in IJCAI, JAIR. ✓ AUTHORIZED TO WORK IN THE US · NO SPONSORSHIP REQUIRED

View Experience Get in Touch
AVAILABLE FOR OPPORTUNITIES
BELLEVUE, WA
SYSTEM_READY // v2026.1
01

About

Dianmu Zhang is an ML engineer with deep experience building and shipping AI systems across agentic AI, computer vision, and on-device ML. She brings a rare combination of research depth — a PhD from the University of Washington, patents, and peer-reviewed publications — alongside strong product execution.

She has led impactful projects at Microsoft, Biorender (Y Combinator), and Optum, including real-time inking prediction used by billions and multi-agent LLM systems for healthcare. Published in recognized AI venues (IJCAI, JAIR) and serves as a reviewer for IEEE RA-L and ICRA.

5+
Granted & Pending Patents
1B+
Users Reached
11
Publications
10+
Yrs Research & Eng
02

Skills

AI / ML

Machine Learning Deep Learning Agentic AI LLMs GenAI Reinforcement Learning Computer Vision

Model Ops & Edge Deployment

Model Training Fine-tuning Model Optimization Edge Inference Shipped on Windows & Android GPU / NPU / CPU ONNX TFLite

Platforms & Languages

Python C# MCP

Machine Learning Frameworks

TensorFlow PyTorch
03

Experience

Optum
2025 — PRESENT
// LEAD AI/ML ENGINEER · BELLEVUE, WA (REMOTE)
  • Engineered multi-agent AI systems (LLM) for patient scheduling and physician decision-support.
  • Implemented advanced reasoning patterns: Chain-of-Thought, Actor-Critic, Reflection.
  • Built grounded retrieval pipelines via MCP.
  • Currently hold 5 pending patents for novel agentic architectures.
Read More →
Biorender (Y COMBINATOR)
2024 — 2025
// SENIOR MACHINE LEARNING SCIENTIST · REMOTE
  • Architected and deployed a multi-stage text-to-figure system for scientific illustration.
  • Utilized LLMs to transform natural language into executable layout code.
  • Successfully shipped to 100% of the Biorender user base.
Read More →
Graham AI
2022 — 2024
// TECHNICAL FOUNDER · BELLEVUE, WA
  • Built AI to distill information from lengthy financial documents.
  • Ranked top 5% portfolio in the Y Combinator selection process.
Microsoft
2018 — 2022
// ML RESEARCHER · REDMOND, WA
  • Built novel neural networks solving problems in Human-Computer Interaction and computer vision — shipped to billions of end users.
  • Inking prediction is one of the first AI models shipped on Windows; led research prototyping to production.
  • Deep learning techniques: embeddings, semantic segmentation, attention-based graph networks, autoencoders, VAEs.
  • Led end-to-end ML development: training from scratch to on-device optimization for Windows and Android (GPU, NPU, tflite, ONNX).
  • Patent granted: Multiple Object Tracking (US 11,017,538).
Read More → The Hero's Journey to Edge Device
University of Washington
2015 — 2020
// PH.D. RESEARCHER · SEATTLE, WA
  • PhD in Electrical & Computer Engineering, focus on AI applications in robotics.
  • Built AI to solve high-cognitive problems: symbolic inverse kinematics via Behavior Trees.
  • Research on reinforcement learning for long-horizon tasks.
  • Published in JAIR, IJCAI, and IROS.
04

Research

01
Multi-Agent Query Resolution Ecosystem For Fragmented Web Domain PATENT
Zhang D, Morato M, Yang W, Hassan S, Mekic I · 2026 · App. No. 19/567,834 · Pending
02
Multi-Agent Web Navigation & Functionality Scraping Agents For Fragmented Web Domain PATENT
Zhang D, Morato M, Yang W, Hassan S, Mekic I · 2026 · App. No. 19/577,738 · Pending
03
Actor-Critic Agentic Input Scheme For Fragmented Web Domain PATENT
Zhang D, Morato M, et al. · 2026 · App. No. 19/630,767 · Pending
04
Multi-Agent Web Scraping and Validation Scheme for Fragmented Web Domain PATENT
Zhang D, Morato M, et al. · 2026 · App. No. 19/577,746 · Pending
05
Multiple Object Tracking PATENT GRANTED
Zhang D, Zharkov I · Grant No. 11,017,538 · May 2021 · justia.com
06
IKBT: Solving Symbolic Inverse Kinematics with Behavior Tree JAIR
Zhang D, Hannaford B · Journal of Artificial Intelligence Research · 2019 · DOI
07
IKBT: Solving Symbolic Inverse Kinematics with Behavior Tree IJCAI
Zhang D, Hannaford B · IJCAI 2020 · Yokohama, Japan
08
Autonomous Inverse Kinematics Solver IROS
Zhang D, Hannaford B · IROS 2017 · Vancouver, Canada

Full publication list on Google Scholar ↗

05

Contact

Let's
Connect

Open to senior ML/AI engineering roles, research collaborations, and speaking engagements.

EMAIL [ CLICK TO REVEAL ]
LINKEDIN Connect PUBLICATIONS Google Scholar
LOCATION Bellevue, WA
✓ AUTHORIZED TO WORK IN THE US · NO SPONSORSHIP REQUIRED