⚙️ LIVE DEMO ACCESS CREDENTIALS:
To test the live Streamlit platform, please use the following pre-configured team profile: • Username: nexus • Password: Nexus_Team@uet
If prompted for administrative access: • Admin Authentication Key: Nexus_Admin_2026
Inspiration
What it does## Inspiration
We are the two founders of Team NEXUS, a robotics team that has won 4 National Championship titles across Sumo Robot, RC Car, and Robo Soccer competitions. Despite our success, our engineering workflow was far from organized. Inventory was tracked manually, tasks were assigned through WhatsApp, and critical technical knowledge was scattered across different files and devices. Before one national competition, we discovered that important components were out of stock only days before the event. At the same time, multiple team members were unknowingly working on the same tasks. We realized that robotics teams needed more than spreadsheets and task boards—they needed an intelligent engineering assistant that understood hardware, software, deadlines, and team operations. That challenge inspired us to build NEXA, an AI Lead Engineer capable of managing real engineering workflows through natural conversation.
What it does
NEXA is an AI Lead Engineer built specifically for robotics teams. It manages live inventory, assigns and tracks engineering tasks, balances team workloads, retrieves technical documentation, and generates competition-ready ESP32 robot code. Through a simple conversation, team members can check hardware stock, view project status, access wiring information, retrieve robot specifications, and generate code for robotics modules. NEXA maintains a centralized knowledge base for our championship-winning Sumo Robot, RC Car, and Robo Soccer platforms. Unlike traditional chatbots, NEXA can perform real actions on live team data, helping robotics teams stay organized and efficient throughout the entire development cycle.
How we built it
NEXA was built using Google Agent Builder as the orchestration layer. We connected the agent to a FastAPI backend deployed on Google Cloud Run and used MongoDB Atlas as the live database for inventory, tasks, and robot documentation. The agent uses tool calling to interact with real-world data and execute actions such as inventory checks and task assignments. The frontend was developed using React and deployed on Streamlit for accessibility across devices. We also integrated our championship robot knowledge base, including hardware specifications, component inventories, wiring configurations, and software modules. The complete system allows natural language conversations to trigger real engineering workflows and database operations.
Database & Memory Layer (MongoDB Atlas Track)
- Persistent Diagnostic History: Integrated MongoDB Atlas to handle unstructured hardware telemetry logs.
- Agent Memory: Used to store and retrieve historical fault resolutions, allowing the Agentic AI to learn from previous IBT-2 driver over-current patches and optimize its debugging protocols over time.
Challenges we ran into
One of the biggest challenges was building a reliable agentic workflow where the AI consistently selected the correct tools and executed them in the right sequence. Complex requests often required multiple database operations and reasoning steps. We also faced challenges designing a database schema that could efficiently support natural-language queries while remaining scalable. Integrating external tools through Google Agent Builder required careful testing to maintain context and ensure accurate responses. Another challenge was transforming years of robotics experience and documentation into a structured knowledge base that the AI could effectively use.
Accomplishments that we're proud of
We successfully built an AI system that functions as a real engineering teammate rather than a simple chatbot. NEXA can manage live inventory, coordinate team activities, generate robotics code, and access technical knowledge in real time. We are especially proud that the project was inspired by real problems faced by our own 4-time National Champion robotics team and was developed into a practical solution. The platform centralizes years of engineering knowledge and makes it instantly accessible to team members. Most importantly, we demonstrated how AI agents can support engineering teams through real-world actions, not just conversations.
What we learned
Building NEXA taught us that successful AI agents depend as much on system design and tool integration as they do on language models. We learned how to design effective agent workflows, structure data for AI retrieval, create scalable APIs, and integrate cloud services into a production-ready application. We also gained valuable experience in prompt engineering, database architecture, agent orchestration, and deploying AI-powered systems that interact with live operational data. Most importantly, we learned how AI can enhance collaboration and productivity in highly technical environments.
What's next for NEXA
NEXA started as a tool for NEXUS Robotics - but the infrastructure we built is designed to scale far beyond one team. Opening NEXA to every engineering team. Our next step is turning NEXA into a platform any robotics club, hackathon squad, or university project group can deploy instantly. Each team gets their own MongoDB instance, their own module library, and their own AI Lead Engineer trained on their specific hardware and competition schedule.
A cross-team community and solo member system. We're building individual engineer profiles that exist independently of any single team. One person can be part of multiple projects simultaneously. NEXA tracks their tasks and workload across all of them. Solo builders can spin up their own NEXA workspace and later merge into a team when collaborators join. Team leads can search the community for engineers by skill set and invite them directly into a project.
Smarter competition preparation. NEXA will auto-generate build timelines counting down from competition day, send procurement alerts before stock runs out, and flag schedule risks before they become crises the exact problem that inspired us to build this in the first place.
GitHub integration. Code generated by NEXA goes directly into the team's repository. No more copying from a chat window. NEXA learns from merged improvements and uses them as the new baseline for future code generation. The goal is simple: give every engineering team at any level, anywhere an AI Lead Engineer that never forgets a component, never loses a task, and always knows what needs to happen next.
Log in or sign up for Devpost to join the conversation.