Hi there, I'm Jinbao 👋

Algorithm Engineer · Recommender Systems · LLM Systems · Quantitative ML

🧠 About Me

• 🎯 Building large-scale recommender systems and LLM-powered applications
• ⚙️ Experienced in TensorFlow production pipelines from training to serving
• 🚀 Passionate about real-time incremental learning and high-throughput systems
• 📈 Exploring the intersection of Quantitative Trading + AI Engineering

Engineering first. Scalability and reliability over flashy prototypes.

🏗 Core Expertise

Recommender Systems

• DSSM / YouTubeDNN / DeepFM / Wide & Deep
• In-batch negative sampling and full-softmax training
• Real-time recall and ranking architecture
• Online incremental training with evolving feature space

Large Language Models

• Transformer architecture implementation and optimization
• LLaMA-style model reproduction in TensorFlow
• Streaming machine translation systems
• Tokenization and inference pipeline design

ML Infrastructure

• Dynamic embedding systems
• End-to-end feature engineering pipelines
• TensorFlow Serving deployment and performance tuning
• Checkpoint compatibility during model/vocabulary expansion

🛠 Tech Stack

• Languages: Python, SQL
• ML: TensorFlow, custom training loops, distributed strategy
• Data & Infra: Hive, real-time feature pipelines, high-throughput embedding services

📊 GitHub Analytics

🚀 Current Focus

• Real-time attention-based lookalike modeling
• Incremental DeepFM with dynamic vocabulary management
• Streaming MT system optimization for low-latency inference
• 0DTE quantitative strategy research with robust risk controls

📌 Engineering Principles

• Production-grade > demo-grade
• System design > model stacking
• Stability and observability > leaderboard overfitting

📫 Contact

• GitHub: @JinbaoSite
• Open to collaboration on recommender systems, LLM infra, and applied quantitative ML.