About Me
Hello👋! I am Yiqiao Jin (靳轶乔, Ahren), a final-year Computer Science Ph.D. candidate at Georgia Tech, advised by Prof. Srijan Kumar.
My research sits at the intersection of multimodal foundation models, intelligent agents, and spatial/physical intelligence: building general-purpose models that can perceive, reason, learn, and act in interactive virtual and physical environments. I work across four threads — vision-language and multimodal models (VLMs/MLLMs); world models, embodied AI, and spatial reasoning; LLM agents and multi-agent systems; and post-training, reinforcement learning, and self-improvement. My work has appeared at ACL, EMNLP, ICML, NeurIPS, ICLR, KDD, AAAI, CIKM, and The Web Conference, including several oral presentations.
Recent work includes AgentArk (NeurIPS 2026), which distills multi-agent reasoning into a single LLM; MASCOT (EMNLP 2026 Main), on persona consistency and collaborative dialogue; and UniSD (COLM 2026 LLA Workshop), on self-distillation. My position papers examine consistency in unified multimodal models and a roadmap beyond structural unification (NeurIPS 2026 Workshop on Linguistic Principles for Foundation Models).
At Amazon, I develop joint policy and world model networks for long-trajectory agentic tasks. Previously at Waymo, I developed a unified VLM for embodied AI perception and spatial reasoning (patent pending). Earlier, I built agentic and multimodal systems at J.P. Morgan AI Research (SlideAgent, ACL 2026 Main; MM-BizRAG, ACL 2026 Industry), Visa Research (SARA, ACL 2026 Main, Oral), Adobe Research, and Microsoft Research Asia. Before Georgia Tech, I worked with Prof. Yizhou Sun and Prof. Wei Wang at the UCLA Scalable Analytics Institute (ScAi).
Selected honors include the MLCommons ML and Systems Rising Stars (2026), Best Paper Award at Good-Data @ AAAI 2025 Workshop, Roblox Graduate Fellowship Finalist (2024), and the Microsoft Research “Star of Tomorrow” Award. My work on cross-lingual LLM evaluation has been featured by Scientific American, The World, and Georgia Tech News.
- Vision-Language & Multimodal Models (VLMs/MLLMs)
- World Models, Embodied AI & Spatial Reasoning
- LLM Agents & Multi-Agent Systems
- Post-training, Reinforcement Learning & Self-Improvement
Ph.D. in Computer Science
Georgia Institute of Technology (Georgia Tech)
B.S. in Computer Science
University of California, Los Angeles (UCLA)
Experience
Applied Scientist Intern
Amazon- Develop joint policy and world model networks for long-trajectory agentic tasks.
- Mentor: Anwesan Pal. Manager: Narayanan Sadagopan.
Ph.D. Intern
Waymo- Developed a unified VLM for embodied AI perception across diverse weather, geographic regions, and lighting conditions (patent pending).
- Built perceptual capabilities for vision-language-action (VLA) systems, including dense depth, object distance, driving relevance, depth ordering, and time-to-reach reasoning, through a shared geometry-language representation.
- Mentor: Prasanna Krishnasamy. Managers: Ming Zou, Christian Lauterbach.
Research Scientist Intern
J.P. Morgan AI Research- Led SlideAgent, a hierarchical agentic framework for multi-page visual document understanding (ACL 2026 Main; patent pending).
- Built multimodal pipelines for slide decks and financial reports, including MM-BizRAG (ACL 2026 Industry Track), integrating context compression and retrieval-augmented generation.
- Mentors: Rachneet Kaur, Zhen Zeng. Managers: Sumitra Ganesh, Manuela Veloso.
Research Collaborator (Visa Research Grant)
Visa Research- Led SARA, a selective and adaptive RAG framework with context compression (ACL 2026 Main, Oral).
- Developed a hybrid compression strategy combining fine-grained spans with compact semantic vectors, achieving consistent gains across Mistral, Llama, and Gemma.
- Mentors: Vineeth Rakesh Mohan, Yingtong Dou, Menghai Pan. Manager: Mahashweta Das.
Research Scientist Intern
Adobe Research- Built ScreenLLM, a multimodal LLM for GUI understanding and action prediction (WebConf'25 Companion).
- Designed a stateful screen schema that summarizes dynamic UI sessions as time-aware textual context, plus a key-frame extractor for significant UI transitions.
- Mentors: Gang Wu, Yu Shen, Stefano Petrangeli. Managers: Saayan Mitra, Vishy Swaminathan.
Research Scientist Intern
Microsoft Research Asia (Social Computing Group)- Mentors: Xiting Wang, Jindong Wang, Xing Xie.
- Published papers across LLMs (ICML'24, ICML'23, AAAI'23), LLM agents (EMNLP'24, ICML'24), misinformation detection (KDD'22, AAAI'22), few-shot learning (ACL'24, AAAI'23), and explainable AI (AAAI'22).
- Received Microsoft Research “Star of Tomorrow” Award (2021).
Software Engineer Intern
Amazon (Fulfillment By Amazon)- Designed and implemented IAR Manual Analysis, a scalable workflow on AWS Step Functions and Lambda that automates aggregation of S3 and DynamoDB data for SageMaker training, handling >16k requests per summary stage.
- Deployed the workflow across all AWS regions (EU/FE/NA) via CloudFormation; built DataCraft pipeline for ingesting DynamoDB tables into the Andes catalog.
Software Engineer Intern
IBM China Development Laboratories- Built Compass DataRouter (Go + MongoDB) for the Compass project, reducing memory usage and accelerating data retrieval.
- Improved the Compass monitoring dashboard with React.js.
Education
Ph.D. in Computer Science
Georgia Institute of Technology (Georgia Tech)- Advisor: Prof. Srijan Kumar.
- Research on multimodal foundation models, world models and spatial reasoning, LLM agents and multi-agent systems, and post-training and self-improvement.
- Publications include NeurIPS'26, EMNLP'26, ACL'26, ICLR'26, ICWSM'26, EMNLP'25, ACL'25, WebConf'24, ACL'24, KDD'23, CIKM'24, AAAI'23.
- GPA: 4.0/4.0.
B.S. in Computer Science
University of California, Los Angeles (UCLA)- Advisors: Prof. Yizhou Sun, Prof. Wei Wang.
- Published 4 papers at top-tier ML and data mining venues (AAAI, KDD, Web Conference). Continuing collaborator on LLMs and graph neural networks (GNNs) (EMNLP'25, WebConf'23).
- GPA: 3.82/4.0. Dean’s Honor List (5 times).
Representative work across vision-language and multimodal models, world models and spatial reasoning, LLM agents, and post-training. The full publication list includes all venues.
📍 Location
Atlanta, GA (during the academic year) — currently in Santa Clara, CA (Amazon internship, Aug–Nov 2026)
yjin328[AT]gatech.edu
🏢 Office
756 W Peachtree St NW, Atlanta, GA 30308
CODA 13th Floor
🕒 Office Hours
Monday - Sunday 9:00 to 20:00