Weijun Luo, PhD

📍 Toronto, ON, Canada
🔗 GitHub | ORCID | Google Scholar | Linkedin

🎓 Education

Georgia Institute of Technology
M.Sc., Computer Science (Machine Learning Track)
Atlanta, GA, USA | 09/2024 – 08/2026

Boston University
Ph.D., Physical Chemistry
Boston, MA, USA | 09/2017 – 05/2023

University of British Columbia
M.A.Sc, Materials Engineering
Vancouver, BC, Canada | 01/2015 – 05/2017

Fudan University
B.Sc, Applied Physics
Shanghai, China | 09/2009 – 07/2013

💼 Work Experience

Scale AI, USA.
Applied ML Engineer — 01/2026 – Present

I. Research at GenAI Applied ML @ Scale AI:

I authored and co-authored ScaleLab’s blogpost:

  1. First-author and core developer of Process-verifier / Reward-hacking measurement tool (Manuscript ready for submission).
  2. Insights Generator: Automated Failure Mode Analysis for Agent
  3. HiL-Dynamics: Understanding Agents That Don’t Know What They Don’t Know. I developed HiL-Dynamics tool
  4. [Who Grades the Graders? Rethinking Verifier Design for Computer-Use Agents]

II. Product development at GenAI Applied ML @ Scale AI

At Scale AI’s GenAI org and Applied ML team, I primarily focus on data quality eval, failure model analysis, and scale up for autonomous quality control. The followings are the projects that I led and completed.

  1. Developed automate-hillclimb package for automated optimization of quality checks against production data; Led real-time / post-completion checks development and training

  2. Devleoped Video-OCR package for OCR+LLM transcription of text from videos

  3. Built end-to-end pipeline for evaluating contributor justification quality for universe generation.

  4. Developed a container-free pipeline for running SWE-Bench Pro tasks with Mini-SWE-Agent: collecting agentic trajectories, and evaluating them with LLM judge via Inspect AI, including a Streamlite web app with background job execution, live trajectory streaming, and job history tracking.

  5. Built Hil-Dynamics (https://github.com/melfeki-11/HiL-Dynamics; https://arxiv.org/abs/2604.09408): it measures how frontier agents behave when a task is underspecified.

  6. Major developer of Scale’s Lighthouse / Insights platform for benchmark and delivery data analysis and insights generation: led the development of failure mode analysis.

  7. Major developer of AgenticDq platform for on-demand and automated generation of quality checks for production queues.

  8. Built an in-house package for enterprise data anonymization and redaction.

  9. First-author and core developer of Process-verifier / Reward-hacking measurement tool (Manuscript ready for submission).

  10. Builder of E2E autopatcher for benchmark loopholes/hackable surfaces and production data for delivery

Scale AI, USA.
AI Training Consultant — 03/2025 – Present
Develop PhD-level STEM / research datasets to advance next-generation LLMs that can drive rigorous, high-impact scientific discovery.

Scale AI, USA.
Queue Manager — 10/2023 – 03/2025
Project lead for multiple reasoning projects in STEM fields

Mayson Science Inc. (Toronto, ON, Canada)
Founder — 02/2025 – Present

ThinkCX Inc. (Vancouver, BC, Canada)
Data Scientist / Data Engineer (AWS Certified) — 02/2023 – 12/2024
Designed and developed a population-wide wireless churning solution

Signify (Philips Lighting) Research North America. (Burlington, MA, USA)
Machine Learning Research Intern — 05/2022 – 08/2022

Boston University. (Boston, MA, USA)
Research Fellow — 09/2017 – 03/2023

  1. Published six first-author research articles with average IF = 11;
  2. Developed SpecSmartDecov: An Automated Highthroughput Optial Spectra Analysis Matlab Package

Oak Ridge National Laboratory (Oakridge, TN, USA)
Visiting Researcher, CNMS — 10/2021 – 11/2022

University of British Columbia. (Vancouver, BC, Canada)
Research Assistant — 01/2015 – 04/2017
Dopant diffusion in strained SiGe HBTs

📚 Selected Publications

  1. Weijun Luo, et al. “Improving Strain-localized GaSe Single Photon Emitters with Electrical Doping.” Nano Letters 23, no. 21 (2023): 9740–9747. (IF=11)
  2. Weijun Luo, et al. “Imaging Strain-Localized Single-Photon Emitters in Layered GaSe below the Diffraction Limit.” ACS Nano (2023). (IF=18)
  3. Weijun Luo, et al. “Deterministic Localization of Strain-Induced Single-Photon Emitters in Multilayer GaSe.” ACS Photonics 10, no. 8 (2023): 2530–2539. (IF=7)
  4. Tan, Qishuo, Weijun Luo, et al. “Charge-transfer-enhanced d–d emission in antiferromagnetic NiPS3.” Applied Physics Reviews 9, no. 4 (2022): 041406. (IF=19)
  5. Chen, Zhuofa, Weijun Luo, et al. “Charge Separation in Monolayer WSe2 by Strain Engineering: Implications for Strain-Induced Diode Action.” ACS Applied Nano Materials 5, no. 10 (2022): 15095–15101. (*: equal contribution) (IF=7)
  6. Weijun Luo, et al. “Excitation-Dependent Anisotropic Raman Response of Atomically Thin Pentagonal PdSe2.” ACS Physical Chemistry Au 2, no. 6 (2022): 482–489. (IF=3)
  7. Weijun Luo, et al. “Anisotropic phonon response of few‐layer PdSe2 under uniaxial strain.” Advanced Functional Materials 30, no. 35 (2020): 2003215. (IF=20)
  8. Tianshu Li, Weijun Luo, et al. “Probing the domain architecture in 2D α‐Mo2C via polarized Raman spectroscopy.” Advanced Materials 31, no. 8 (2019): 1807160. (IF=30)
  9. Weijun Luo, et al. “Thermal sublimation: a scalable and controllable thinning method for the fabrication of few-layer black phosphorus.” Nanotechnology 28, no. 28 (2017): 285301. (IF=4)

🎤 Conference Talks

  1. “Electrically Tunable Brightness and Single Photon Purity in Multilayer GaSe.” — CLEO (Conference on Lasers and Electro-Optics, Optical Society of America Annual Meeting), 2023
  2. “Voltage-controlled GaSe single photon emitters.” — APS (American Physical Society Annual Meeting), 2023
  3. “Near-deterministic generation of defect-related emitters in GaSe.” — APS, 2022
  4. “Probe the anisotropic Raman response of atomically-thin pentagonal PdSe2 with 13 excitation wavelengths.” — ACS (American Chemical Society Annual Meeting), 2022
  5. “Thermal sublimation: a scalable and controllable thinning method for the fabrication of few-layer black phosphorus.” — MRS (Materials Research Society Annual Meeting), 2016

💰 Funding

🏅 Awards