Shao-Qian (SQ) Mah

- Core researcher on Codex, GPT-5 Codex and onwards - GPT 5 - GPT 4.5

Building the best LLM tutor for learners and teachers. Responsible for model alignment, evaluation, and human data.

Google Play Growth infrastructure

The world's easiest and most accessible way to learn AI. No code, no math necessary. Find out more at aiblocks.org.

Learn more at getaiblocks.com. • Created a computer vision model to read train location, train track configurations and diagnostics to predict and prevent train collisions in the Hong Kong Mass Transit Railway (MTR), one of the busiest metro systems in the world, serving 1.3+ billion passengers per year • Developed an automatic license plate recognition system that can process multiple video feeds deployed in 80 locations around Hong Kong from scratch with Python and Tensorflow • Constructed data collection, processing and autoscaling ML inference pipelines that streams data from 200+ cameras with Python, Kubernetes and Flask

AI+Cloud Product Manager • Worked across design and software engineering to spearhead the development of a multi-modal data and AI platform to accelerate AI model development and deployment • Architected computer vision model metadata and pipeline to enable platform and ML infrastructure-agnostic execution of computer vision model pipelines • Collaborated with design to create drag and drop graph-based UX to chain different computer vision models together • Conducted customer research to understand their painpoints with AKS-HCI, synthesized in a competitive analysis with Google Anthos presented to product leadership

• Worked on explainable AI (XAI) model architectures, including Transformer-based architectures, neural backed decision trees and deep reinforcement learning. • Investigated explainable AI techniques (LIME, Shapley, local activation, attention) for finance, high frequency trading (HFT) and medical imaging.

• Led and recruited team of 12 from business, design to machine learning • Used React Native and Redux hybridized with native Swift code to write cross-platform iOS, Android and Web application • Wrote Flask backend using Python and SQL to fetch and respond to API requests from the mobile app and web platform, and to power a personalized recommendation algorithm • Winner of 2018 Entrepreneurs@Berkeley pitch competition; speaking with VCs for potentially 200,000 USD in funding

• Led the initiative to enable AI/ML time series forecasting and anomaly detection for business KPIs across the Oracle Analytics Cloud platform • Worked with $40 billion+ revenue customers to understand their forecasting pain points, created vision for the product with mockups in Sketch • Collaborated with development team to deliver an end-to-end predictive analytics and anomaly detection pipeline using Python • 1 of 11 summer PM interns

Hydropal is a startup promoting a healthy lifestyle with a smart, water-intake tracking water bottle • Created microcontroller and sensor hardware design, iOS companion app and Bluetooth communication protocol to enable the smart water bottle to track how much water the user drank, and send the information to their phones • Led a team of 26 through the full design, prototyping and manufacture of the water bottles from scratch • Developed the iOS companion app in Swift and published it onto the App Store; wrote water bottle software in C • Negotiated volume order discounts for electronic components with suppliers and sold 100+ of the water bottles

• Used React and threejs to construct VR solar system on the web with on-the-fly orbit calculations based on real Keplerian orbital elements and physics • Launched the VR solar system as part of demonstration to schools to pursue VR education • Represented IOIOCreative at startup conferences to secure funding for future R&D projects

Awarded $127,000 merit-based scholarship

$140,000 merit-based scholarship Full tuition covered for 7 years