Inspiration
Enterprise IT help desks are notoriously bottlenecked by repetitive triage tasks, lost device asset tracking, and unauthorized access vulnerabilities. We were inspired to build NovaDesk after recognizing that while standard AI chatbots can chat, they lack the secure, enterprise-level integration required to actually solve corporate hardware issues. We wanted to create a context-aware assistant that seamlessly bridges the gap between secure employee databases and automated technical support.
What it does
NovaDesk is an intelligent, secure IT Help Desk Agent designed for corporate environments. Upon initiation, the agent enforces a strict security guardrail, requiring a valid employee credential to unlock technical diagnostics.
- Unauthorized Users: If an unverified or blacklisted ID attempts to access the system, NovaDesk triggers an immediate database exception and denies access.
- Verified Employees: Once an ID (such as E1402) is authenticated, NovaDesk pulls the employee's corporate hardware profile from the database. It can instantly diagnose specific device symptoms (like high CPU load, thermal overheating, or fan noise) and automatically escalate critical issues by generating structured technical support tickets.
How we built it
NovaDesk was built using a modern, decoupled full-stack architecture optimized for speed and enterprise security:
- Frontend: A responsive web UI built with JavaScript that handles user message streams and passes session parameters.
- Backend: A high-performance FastAPI server written in Python to orchestrate API routing, session states, and security logic.
- Database layer: MongoDB acts as the secure asset register, mapping active Employee IDs to specific corporate hardware logs.
- AI Core: Built on top of Google Cloud’s Vertex AI Agent Builder, leveraging advanced system instructions and context injection to keep the agent strictly focused on professional IT triage protocols.
Textual diagram: [User Interface] <---> [FastAPI Backend] <---> [MongoDB Registry] | +---> [Google Cloud Vertex AI (NovaDesk)]
Challenges we ran into
Our biggest hurdle was navigating the integration between the local backend server and the enterprise cloud API infrastructure. We ran into rigid connection errors and a series of 404 Not Found roadblocks due to regional endpoint restrictions (us−west1) and shifting syntax method constraints within the Google GenAI SDK.
When the cloud resource routing threatened to stall our progress close to the submission deadline, we adapted by engineering a robust, state-based session management layer directly into our FastAPI backend. This ensured our data flow logically validated credentials and routed user text to the correct operational tracks seamlessly.
Accomplishments that we're proud of
Bulletproof Guardrails: We successfully built an explicit security gate that prevents unauthenticated users from accessing corporate diagnostic workflows. Dynamic State Handling: We created a logical, step-by-step conversational flow that mimics a real-world enterprise application rather than a basic text-in, text-out script. Resilience Under Pressure: Pivoting our backend architecture under an incredibly tight hackathon timeline to ensure a flawless, working presentation build.
What we learned
This project was an intense masterclass in full-stack architecture, database state management, and the realities of deploying enterprise cloud APIs. We learned how critical it is to design strict order-of-operations in backend routing to avoid logic conflicts. Most importantly, we learned how to stay agile, troubleshoot complex library errors under immense pressure, and engineer creative programmatic fallbacks to deliver a highly professional prototype.
What's next for NovaDesk | IT Help Desk Assistant
The next phase for NovaDesk involves integrating live system API hooks to dynamically fetch real-time task manager telemetry directly from an employee's laptop. We also plan to implement full CRUD operations with our MongoDB database so NovaDesk can write newly generated ticket structures directly back into the IT administrative queue, creating a completely automated end-to-end support pipeline.
Built With
- css3
- fastapi
- gemini
- google-cloud
- html5
- javascript
- mongodb
- node.js
- python
- uvicorn
- vertexaiagentbuilder
Log in or sign up for Devpost to join the conversation.