Tong Zeng Email & Phone Number

Blacksburg, Va, Us

Blacksburg, Virginia, United States

• Led a project on autonomous driving scene understanding using Large Vision Language Models (VLM), fine-tuned VLM to enhance traffic concept alignment, scene text recognition, position, and distance awareness, and video captioning, resulting in improved perception and reasoning in complex driving scenarios.• Designed a comprehensive dataset and benchmark for evaluating visual language models in autonomous driving, and developed an open arena for VLM performance comparison and ranking, facilitating better assessment and improvement of model capabilities.• Implemented a multi-agent-based video captioning system to advance the interpretation of traffic incidents (crash and near-crash scenarios), which aimed to improve the safety of autonomous driving systems by enhancing incident analysis.

Ningbo, Zhejiang, China

• Led the architecture design of the Geely Smart Voice Assistant, which supports vehicle control, knowledge-based Q\&A, and chit-chat. Achieved a breakthrough from 0 to 1 in Geely's self-developed Smart Voice System.• Led the design and development of a smart question-answering system using Knowledge Graphs and Retrieval-Augmented Generation (RAG) with Vector Search, Elasticsearch, and Large Language Models. Successfully deployed in three new Geely vehicle models, supporting intelligent Q&A for tens of thousands of cars.• Led an AI Wallpaper Generation Project, utilizing Stable Diffusion to generate wallpapers for in-vehicle infotainment systems and a wallpaper marketplace. Collected and analyzed user preference to train an image scoring model, and deployed an image safety audit model to ensure content safety.• Led cutting-edge research on Human‑AI Interaction (HAX) and Artificial Intelligence Generated Content (AIGC) technologies, resulting in the successful filing of 7 invention patents. Involving voice interaction, multimodal interaction, information processing methods, text generation, image generation, and music generation.• Supported team recruitment and training by establishing interview feedback standards, developing an interview question bank, and designing the training curriculum. Conducted interviews for over 100 candidates ranging from new graduates to senior experts, recruited more than 10 employees, and mentored 2 junior employees.

Shanghai, China

• Led the design and implementation of the server-side webpage rendering system for the JD.com App, responsible for generating H5 pages for core channels and tens of thousands of promotional activities. This system is one of the earliest and largest Node.js services on JD.com.• Designed and implemented DevOps infrastructures for Node.js services, including deployment, monitoring, and auto-scaling, to ensure high availability and fault tolerance during campaign activities and the 11.11 shopping festivals.• Enhanced the internal Node.js application development ecosystem by constructing a private NPM registry and developing over 5 Node.js SDKs for logging, system monitoring, Redis caching, file storage, etc. Promoted Node.js adoption by sharing experiences through internal and public presentations.• Executed the distributed deployment of a system with thousands of CPU cores, completing the transition from client-side to server-side rendering within one year. Improved first screen rendering time from 3 seconds to under 1 second, supporting tens of thousands of channels and promotional pages with hundreds of millions of daily page views and peak QPS in the tens of thousands.

Master Of Science - Ms, Computer Science, 3.9/4.0

Doctor Of Philosophy - Phd, Information Science

Joint Ph.D. Program - Research Training, Information Science