Fei Wang

Own AI technical strategy; build and lead the engineering/ML organization; partner on fundraising. Built a Data Platform-as-a-Service and the MaidX multi-agent orchestration layer for end-to-end post-training data pipelines.

• Built model/data evaluation infrastructure: reward-model training and evaluator stacks to continuously score, select, and improve datasets
• Implemented agentic RL training loops with repeatable offline evaluation and quality gates for iterative improvement
• Drove "Data 2.0" platform shift delivering 80%+ automation of enterprise data services
• Achieved 5× faster pipeline execution and 10–50% cost reduction through disciplined compute optimization

Led the development of foundation models for electromagnetics, acoustics, and fluid dynamics, translating physics AI from research to industrial applications to accelerate innovation.

• Secured $3M seed funding by crafting technical vision and investor narrative, demonstrating executive alignment and business potential
• Led rapid prototyping of core product (GenAI-driven platform), delivering MVP milestones and customer validation to secure early market traction
• Defined company strategy and product roadmap, integrating market segmentation and go-to-market plans with scalable technical architecture
• Built the founding engineering team: recruited and onboarded key engineers, establishing development processes and quality bars to drive velocity

Spearheaded all facets of Prime Video's recommendation and search systems, leading integration of Generative AI with personalization frameworks. Grew team from 10 to 40 scientists and engineers.

• Single-threaded science owner for Prime Video's HomePage and Title Detail Page, driving discovery and monetization
• Led StreamFM, a transformer foundation model powering Prime Video ranking — drove +5 min watch time per user per week
• Launched GenAI-powered "AI Topics": grouping, naming, and dynamic categorization across detail pages
• Led Search Decoupling initiative achieving >10% increase in Search Attributed Streaming Time
• Modernized the ML platform, accelerating algorithm development cycles by 2×

Built and led a high-performing AI research team from 0 to 15 researchers. Delivered strategic updates to C-suite including CEO and CTO.

• Deep Authentication Project: AI fraud detection boosting detection rates 20%, projected $20M new revenue
• Cybersource Recommender System: low-latency (<10ms) system projecting $100M in revenue growth
• Scalable Feature Aggregation: 10× faster processing vs. native Spark/Redis, saving millions in hardware costs
• Produced 20+ pending patents and 15 papers in top-tier conferences

• Led AI-powered smart parking deployments across multiple cities, optimizing urban mobility
• Built end-to-end deep learning pipeline for vehicle/license plate detection, car model classification, and OCR — deployed on edge and cloud

• Launched Amazfit Health Band, the company's first medical-grade product (CFDA clearance)
• Transformed all medical algorithms to deep learning (biometric ID, sleep staging, signal compression) with production performance above SOTA
• Developed disease classification from biosignals (ECG/PPG/GSR) for AFib, PAC/PVC, sleep apnea; prepared CFDA/FDA filing documentation
• Filed 10+ patents based on core assets

• Responsible for first-ever FDA approval of an algorithm detecting Atrial Fibrillation, plus two additional FDA-cleared algorithms
• Established strategic research roadmap and big data strategy — hosting ECG algorithms as an API service
• Filed 10+ patent applications based on core company assets

• Shipped social computing platforms to IBM Watson/BlueMix: PEARL (emotion analysis), Smart SocialQA, and REACH (relationship analytics)
• Co-led "Medical Sieve" radiology anomaly detection — winner of IBM Research Grand Challenge; integrated into IBM Watson Health Imaging
• Built "AALIM" multimodal healthcare analytics for Kaiser Permanente; deployed "Clinical Data Hub" at Cedars-Sinai (16,000+ patients)