This project develops a framework that breaks down LLM outputs into clear, modular components, making them easier to understand, edit, and reuse. By enabling structured and adaptable responses, it improves clarity, flexibility, and collaboration in AI-assisted work.This project is building a flexible, hardware-agnostic software ecosystem for embedded systems using containerization, cloud tools, and peer-to-peer communication. It enables scalable, over-the-air updates and decentralized computing across diverse devices.
‍
99P Labs Team:
• Ryan Lingo
• Martin Arroyo
• Luka Brkljacic
• Rajeev Chhajer
• Ben Davis
• Nithin Santhanam
‍

This project is building a flexible, hardware-agnostic software ecosystem for embedded systems using containerization, cloud tools, and peer-to-peer communication. It enables scalable, over-the-air updates and decentralized computing across diverse devices.
‍
99P Labs Team:
• Brian Coy
• Ben Davis
• Luka Brkljacic
‍

The LLM Landscape Report by 99P Labs provides a comprehensive overview of large language models (LLMs), their industry impact, research initiatives, and technological advancements. It covers key trends such as the rise of Retrieval-Augmented Generation (RAG), small model experimentation, and the integration of LLMs into various applications. The report highlights 99P Labs’ research contributions, including methods like REAP for structured problem-solving and student innovation projects exploring AI applications in capstone programs. It also touches on best practices for working with LLMs, including prompt engineering and infrastructure considerations, providing insights for researchers, businesses, and AI practitioners looking to leverage LLMs effectively.
‍
99P Labs Team:
• Ryan Lingo
• Martin Arroyo
• Rajeev Chhajer‍

This paper develops a system for vehicles to share sensor data in real-time, enhancing perception, safety, and traffic efficiency. By optimizing data sharing and processing, it enables smarter, more connected transportation with future potential for real-world deployment
‍
99P Labs Team:
• Nithin Santhanam
• Rajeev Chhajer

UCSD Team:
• Sam Thornton
• Sujit Dey
‍

PlantD is an open-source tool for analyzing and optimizing data pipelines by orchestrating synthetic data generation, measuring cost, latency, and throughput, and managing load patterns. It provides clear metrics and visualizations to help developers quickly assess different pipeline configurations for better efficiency and reliability.

99P Labs Team:
• Rajeev Chhajer
• Brian Coy
• Ben Davis
• Tony Fontana

CMU Team:
• Majd Sakr
• Chris Bogart
• Baljit Singh

99PLabs-F1Tenth is an open-source project built around a 1/10th-scale autonomous race car platform, enabling rapid development and testing of autonomous vehicle software. Through modular Docker-based architecture and simulation tools, it accelerates innovation and safe deployment in both virtual and real-world settings.

99P Labs Team:
• Brian Coy
• Nithin Santhanam
• Ben Davis
• Luka Brkljacic
• Tony Fontana
• Martin Arroyo
• Rajeev Chhajer
• Ryan Lingo
‍

REAP is a method that combines reflection, explicit task decomposition, and advanced prompting to enhance Large Language Models for complex, multi-step tasks. It improves coherence, clarity, cost-efficiency, and scalability by iteratively refining solutions and exploring multiple pathways to better
answers.
‍
99P Labs Team:
• Ryan Lingo
• Martin Arroyo
• Rajeev Chhajer

This research investigates daily traffic pattern changes resulting from significant network disruptions using data-driven incremental Non-negative Matrix Factorization (NMF). Employing high-resolution traffic speed data collected from the I-95 corridor in Pittsburg, the study analyzes how spatial-temporal traffic features evolve during normal and disrupted states. By integrating multi-stage incremental learning with route-based and topological regularization methods, this work identifies stable and adaptive spatial features reflecting dynamic traffic behaviors under disruptions. Preliminary results demonstrate distinct temporal shifts in traffic peaks following disruptions, revealing key routes' resilience and vulnerability.
‍
99P Labs Team:
• Nithin Santhanam
‍
HRI-EU Team:
• Sebastian Brulin
‍
CMU Team:
• Jingyan Li,
• Sean Qian

This study presents a traffic forecasting model that predicts travel times several hours ahead and reports how confident it is in each prediction. Its main idea is a road entwork model that changes hour by hour and reacts to crashes, weather, and work zones, instead of assuming conditions stay fixed. Tested on Ohio traffic data and Columbus simulations, the model beat existing method on long-range accuracy and gave reliable uncertainty ranges.
‍
99P Labs Team:
• Allen Sheldon
• Rajeev Chhajer
• Nithin Santhanam
‍
OSU Team:
• Mayur Patil
• Qadeer Ahmed
• Shawn Midlam-Mohler

Ohio Department of Transportation Team:
• Stephanie Marik