Sandvik Coromant is a world-leading supplier of tools and tooling systems for metal cutting. Behind that sits a production environment with a large installed base of automation equipment – some of it still running on Adept robot controllers programmed in V+. 

This thesis project covers AI as an engineering aid when modernizing legacy Adept/V+ robot installations. 

This is a proof-of-concept project with honest premises. We do not believe AI can replace robot, PLC, and safety engineers in a control system migration, and we are not asking you to prove that it can. What we want to know is something more useful: how much of the manual groundwork involved in such a migration can AI take off our hands? 

Background and scope of the project 

Parts of our production still rely on obsolete Adept/V+ robot controllers, while the mechanical equipment has many years of useful life left. Modernization is often postponed because it requires engineers to interpret undocumented code, reconstruct system behavior, and create documentation and tests. Although the architectural differences make automatic translation unrealistic, AI could support this groundwork through analysis, documentation, code drafts, and test generation. 

Purpose 

The purpose of the project is to quantify, on a single small and well-defined part of a real V+ application, how much AI can contribute in four specific areas: 

• Analysis – understanding and explaining existing V+ code, its structure, dependencies and behavior 

• Documentation – producing the functional and technical documentation that is missing today 

• Code drafts – generating first-draft code for the target platform for an engineer to review and complete 

• Test generation – deriving test cases and verification criteria from the original application 

For each area we want a measured answer: how much of the work could AI do, what was the quality, how much rework did an engineer have to put in, and how much time was saved in total? 

What this project is not 

The goal is explicitly not a fully automated, production-ready conversion. Success is a clear, well-substantiated picture of where AI helps and where it does not and a working method we can reuse. A result showing limited benefit is a perfectly acceptable outcome. We are after an honest answer, not a positive one. 

Work description

The project runs as a proof of concept in five phases. All verification is done virtually against a 3D model; no commissioning on physical equipment is involved. 

1. Scoping and reference behavior 

• Together with your supervisor, select one small and well-defined part of a V+ application – a single sub-sequence or program module – as the case for the whole project 

• Establish a measurable reference behavior for that part – motions, I/O, sequence logic and interfaces to surrounding equipment. This is what every later result is judged against 

• Define up front how AI contribution will be measured in each of the four areas, so the comparison is fair and repeatable 

2. AI-assisted analysis 

• Use two or three AI tools to interpret and explain the V+ code: structure, control flow, dependencies and implicit assumptions 

• Assess correctness against the reference behavior – what did the tools understand, what did they miss, and what did they state confidently but wrongly? 

3. AI-assisted documentation 

• Generate functional and technical documentation of the selected part with AI support 

• Evaluate how much of the output is usable as-is, how much needs editing, and how it compares with documenting the same code manually 

4. AI-assisted code drafts and test generation 

• Produce first-draft code for the target platform with AI support, intended for engineering review rather than direct use 

• Generate test cases and verification criteria from the original application 

• Document failure patterns systematically. Where the tools go wrong is at least as valuable a result as where they succeed 

5. Verification against a 3D model, and recommendation 

• Build a simplified kinematic 3D model of the equipment – reusing existing CAD material where it exists – and run the reviewed code against it to verify motions, sequence and I/O against the reference behavior 

• Summarize the measured AI contribution per area, with an estimate of time and cost saved compared with a fully manual approach 

• Recommend how Sandvik Coromant should work with AI support in future modernizations, which steps to use it for, which to keep entirely in engineering hands, and what is needed to scale the method 

Scope and limitations 

• The work covers one small part of one V+ application, not a complete program or machine 

• Verification is virtual. Commissioning on physical equipment is not included 

• The 3D model is kinematic and limited in scope. Physics modelling, cycle time optimization and full mechatronic simulation are out of scope 

• The project covers control and software, not mechanical rebuild 

• Safety functions are analyzed and documented, but safety design and certification remain with our engineers and are not delegated to AI tools in this project 

Location 

In place at Gimo, Sweden. 

Student background 

We are looking for one student with an interest in automation and a healthy skepticism towards tools that produce confident-looking output. 

What you bring: 

• You are in the final year of a Master’s or Bachelor’s degree in automation, mechatronics, electrical engineering, computer engineering, or a related field 

• Basic knowledge of PLC programming and/or industrial robotics 

• Interest in simulation and virtual commissioning (experience with CAD or 3D simulation is an advantage) 

• You work systematically and critically, and are comfortable measuring and documenting results rather than only building things 

• Good communication skills in English or Swedish 

Experience of AI tools for software development is a plus. 

Our culture 

At Sandvik, we’re tech-driven, innovative, and entrepreneurial. We believe that success is a team effort, so we value diversity and are committed to creating an inclusive culture where people can be themselves and reach their full potential. We invest in supporting each other, learning together, and celebrating our differences. Visit our stories hub, LinkedIn, or Facebook to get to know us better. 

Duration of the project 

The thesis project corresponds to 30 credits and lasts approximately 20 weeks, starting at the beginning of 2027. Upon completion and approval of the thesis, Sandvik offers lump-sum compensation. 

Contact information 

For further information about this project, please contact Jacob Bexenius, Engineering Blanks Manager, +46 (0)76-767 39 97. 

We’ve already decided on which advertising channels and marketing campaigns we wish to use, and respectfully decline any additional contacts in that matter. 

How to apply 

We have an ongoing selection process for this project and ask you to send your application as soon as possible, but no later than November 20, 2026. Click Apply and include your resume and cover letter in English. Please note that we don’t accept applications by email. Job ID: R0097667.