sunbc0120 - Overview

Baichuan Sun, Ph.D.

Lead GenAI Engineer @ Google APAC

Orchestrating the Frontier of Enterprise Intelligence & Infrastructure Economics

Bridging Deep R&D, Corporate Strategy, and Production-Grade Engineering

🌐 The Convergence of Intelligence & Strategy

I architect the bridge between scientific rigor and sustainable business value. My work centers on translating the kinetic energy of frontier AI research into the potential energy of enterprise transformation. From computational physics to boardroom decision-making and hyperscale cloud AI engineering, I navigate the full stack of modern intelligence.

🚀 Current Focus: Frontier GenAI & Reasoning

At Google, I advise APAC’s enterprise C-suites on the sovereign adoption of Generative AI, focusing on:

Agentic Data Engines: Moving beyond simple RAG into autonomous, multi-modal reasoning loops.
Infrastructure Economics: Optimizing elastic HPC and training/inference costs for the next billion tokens.
Cognitive Trust: Engineering the guardrails, red-teaming frameworks, and safety layers required for enterprise-grade compliance.
Next-Gen CX: Deploying hyper-personalized, real-time multimodal agents that redefine human-computer interaction.

🏗️ Strategic Engineering & Selected Research

Bridging the gap between frontier research and production-grade systems.

🧠 Reasoning AI (GRPO)

Architected distributed NeMo-RL clusters on GKE to fine-tune Gemma 3 using Group Relative Policy Optimization (GRPO). Optimized FSDP sharding on NVIDIA B200 clusters.

Value: Achieved +6.2% absolute gain in MATH-500 accuracy via custom reward engineering.

🔍 Enterprise Document AI (OCR)

Architected a multi-modal Financial Audit engine using LMMs and advanced OCR. Automated the extraction of complex, unstructured data from heterogeneous financial instruments.

Value: Eliminated 90% of manual data entry for Tier-1 financial institutions.

🤖 Spatial Intelligence & Robotics

Developed high-fidelity Computer Vision pipelines for autonomous robotic systems. Focused on real-time object detection and kinematic path planning in dynamic environments.

Value: Reduced manual oversight by 40% in industrial automation.

⏳ Predictive Prognostics

Leveraging my background in Thermodynamics, I built Time Series Forecasting models for high-value industrial assets to predict failure modes before they occur.

Value: Saved millions in unplanned downtime for energy providers.

🛠️ Technical Ecosystem

Frontier GenAI & Research	Scalable Engineering & HPC	Strategy & Domain
`Gemini / Gemma 3`	`NVIDIA B200`	`C-Suite Advisory & Consulting`
`GRPO & RLHF (NeMo-RL)`	`Ray on GKE (KubeRay)`	`Unit Economics of AI`
`Vertex AI & Model Garden`	`PyTorch FSDP & DCP`	`AI ROI & TCO Modeling`
`Agentic RAG / Agents`	`vLLM & FlashInfer`	`Sovereign AI Frameworks`
`Dynamic Grounding`	`Distributed Training (XLA)`	`Gov & Risk Management`

📈 Global Impact & Thought Leadership

"Science is only as useful as its ability to be democratized."

Lead GenAI Solutions: Architecting AI roadmaps for APAC Decacorns, transforming legacy data into Agentic Intelligence.
Open Source Authority: 80K+ downloads of tools designed to bridge the gap between complex data assets and business value.
3M+ Professionals reached via technical publications and strategic guidance on AI infrastructure and economics.
Global Footprint: Mechanical Engineering (🇨🇳) → Robotics Researcher at Tohoku (🇯🇵) → Statistical Physics PhD at NTU (🇸🇬) → Wolfram Summer School (🇺🇸) → CSIRO/McKinsey/AWS (🇦🇺) → Google.