High Level Description

AI agents connected to real infrastructure via the Model Context Protocol (MCP) or such, can already read live system state and take real action, such as scaling a Kubernetes deployment, restarting a pod, or filing a Jira ticket. What is still an open question is how much autonomy such an agent can safely be given in a production-like environment before the risk of an incorrect action outweighs the benefit of speed. This thesis investigates this question in a sandboxed environment built by the student.

Project Description
The student builds a small, isolated environment (a Kubernetes cluster and a Jira board) and connects AI agents to it via MCP tool calls (or something else). The agent is evaluated across a set of graded autonomy levels, from observing and filing tickets in Jira to proposing actions for approval to acting autonomously within tightly scoped, low-risk operations. Controlled failures (pod crashes, resource exhaustion, network faults) are injected, and the agent’s diagnoses, proposed or executed actions, and generated tickets are compared against a human baseline.

Who are we looking for?

• Bachelor/Master of Science in Computer Science/Engineering

• Graduation in June 2027

• Fluency in Swedish, both written and spoken

• Fluency in English, both written and spoken 
  

Purpose
Bachelor students typically focus on one question; master students take a broader comparative evaluation. In this thesis, investigate these questions:

• How does the level of autonomy granted to an AI agent (observe-and-escalate vs. propose-and-approve vs. autonomous-within-bounds) affect the accuracy and safety of its actions in a simulated Kubernetes/Jira environment?

• How accurately can an LLM-based agent diagnose the root cause of an injected infrastructure fault compared to a human operator, and how is that accuracy measured objectively?

• How useful and accurate are automatically generated Jira tickets (context, root-cause hypothesis, affected services) compared to tickets written by a human on the same incidents?

• What architecture, guardrails and permission model (e.g. Kubernetes RBAC) are needed to safely integrate such an agent into an existing DevOps workflow?


An Exciting Journey with Knightec Group
Semcon and Knightec have joined forces as Knightec Group. Together, we are Northern Europe’s leading strategic partner in product and digital service development. With a unique combination of cross-functional expertise and a holistic business understanding, we help our clients realize their strategies – from idea to complete solution.

Practical Information
This is a thesis position, located at our office in Sundsvall. Start date January or March 2027.

Please submit your application as soon as possible, but no later than 2026-11-30. If you have any questions, you are welcome to contact johanna.edstrom@knightec.se. Note that due to GDPR, we only accept applications through our careers page.