START Hack Vienna 2026 - techbold
Build the AI technician workspace.
techbold runs IT for more than 1,200 businesses across Europe. This track asks teams to build controlled agentic AI for diagnosing and fixing Linux incidents.
Challenge
From ticket to validated fix
Build a technician workspace that pulls assigned tickets from a mock Phoenix ERP, loads the affected customer system, connects over SSH, and diagnoses the incident.
The AI may propose diagnostic and fix steps, but the technician must approve every system action. After validation, the workspace writes a precise activity log back to the ERP.
The evaluation uses fresh systems your team has not seen before, so hard-coded fixes and unsafe broad commands will not hold up.
What to build
Pull tickets from a mock ERP, connect over SSH, diagnose and fix, validate, then write a clean activity log.
Requirements
Submission requirements
- Public GitHub repo
- Up to 3-minute demo video
- Written submission
Judging
How this track is scored
- Functional MVP and ERP workflow (20%): Ticket loading, customer context, activity creation, and robust states work.
- Troubleshooting performance (35%): Hidden incidents are solved with correct root cause, durable fix, and no regressions.
- Safety, auditability, responsible AI (20%): Actions are auditable, controlled by humans, minimal, and secret-safe.
- Technician experience (10%): The operator can understand, review, retry, and abort agent work.
- Engineering quality (15%): The repo is reproducible, modular, documented, and tested where it matters.
Prize
Prize for the winning team
- EUR 3,000 in hardware plus a paid EUR 7,000 contract for a real techbold case
Resources
Resources and support
- A social-media winner story with a techbold board endorsement
- Production-grade infrastructure and Linux systems to build on
- Direct line to techbold's AI engineers on site and on Discord
- Mock Phoenix ERP API: Bearer-token API with endpoints for open tickets, customer system info, and activity creation.
- Five customer Linux VMs: Fresh Ubuntu machines with incidents discovered through assigned tickets.
- SSH key and reset endpoint: Team-specific SSH access and a reset endpoint for replaying cases.
- Safety hard fails: Deleting customer data, broad chmod/chown, disabling security controls, exposing secrets, or hiding logs can zero an incident or disqualify a team.
Mentors / Jury
People in the room
- Christopher Chellakudam: Senior Developer and AI Engineer, techbold - available onsite
- Benedikt Fritzenwallner: Senior Developer and AI Engineer, techbold - available onsite
Mentors / Jury
Mentors for all cases
- Cristian Livadaru: n8n ambassador - mostly available on Day 1
- Gleb, Vlad, Sviat, and the proximata team: Available on the fourth floor in the Robotics Area