This page details the services [0,A,B,C,D,E] I can provide. Below each service I have describe the tasks I expect could be relevant along with my sources of experience.

(0) Initial consultation(s)

I need to understand your problem and what kind of solution you want. This is always helped by examples, pictures and numbers. In some cases a physical/virtual tour could be necessary.

  • I sign your NDA?
  • You explain your problem and wishes
  • We discuss if and how I can help, what is realistic, success criteria and resources
  • I present a plan (equipment, tasks, deliverables, price/hours)
  • We adjust and agree on the plan
  • We negotiate and sign a collaboration agreement. Do you provide a draft or should I?

If the problem is loosely defined or too complex for us to settle on a plan, I suggest we a lot a number of hours of (A).

(A) Idea generation and project proposals

Producing ideas and project proposals is often a balance, we want to set the bar high in order to be able to achieve something impressive. Meanwhile, we also want to ensure that the goals can be reached within the given constraints. Idea generation should include a physical/virtual visit to the location with light data collection.

  • Identify variables, success criteria and metrics, and perform risk assessment
  • Simple illustrations/simulations
  • Idea or project proposal presentation (ppt,pdf)
  • Grant application


  • Worked on 14+ real-world projects
  • 6+ master level student projects I have proposed and supervised.
  • Among the grants I have contributed to are the following: financial support of 1 + 1 + 21 + 0.6 = 23.6 mio. DKK = 3.9 mio. USD)
    • 2x successful industrial Ph.D projects (success rate 50%) and 1x successful Grand Solution project (success rate 10-20%)Innovation Fund Denmark
    • 1x successful Norma og Frode Jacobsens Fond
    • 1x failed Horizon 2020 (success rate 16%) EU
    • 1x failed Novo Nordisk Foundation (success rate 20%) EU
  • Strong network in food industry and computer vision

(B) Background research or training

Report on the current solutions and techniques in a given domain along with suggestions for when one or the other technique should be used. Because of the pace of

  • Presentation of findings, technologies (pros and cons)
  • Hands-on tutorial for teaching the use of a tool


  • Every scientific paper and most grant applications must be placed in the context of existing solutions and techniques, thus background research is a common task as university researcher

(C) Data collection and experimentation

Collection quality data is hard work but hours are often better spend here that tinkering with the Neural Networks. Experience shows that although problems in datasets sometimes can be fixed after the data has been collected, this is time consuming and far from optimal. For this reason to be involved in the collection process and to at least collect a small amount of data first that can be processed in order to reveal problems before the bulk is collected.

  • "Light weight" "hand-held" data collection
  • "quick-n-dirty" analysis of collected data
  • Build data collection setup hardware+software

(D) Second opinion/assist an ongoing project

It can be helpful to have a fresh set of eyes involved in machine learning projects. The nature of machine learning makes it easy to get fooled by the apparent performance of the model. The speed and scope of machine learning is expanding at a blazing pace and techniques from one domain or application are often useful in other areas. This makes it impossible for to one set of machine learning engineers to keep an up-to-date overview of all of the possibilities.

  • Consider choices and propose alternatives
  • Inspect the data processing and system evaluation
  • Stress test the solution
  • Contribute part of the pipeline


  • Depending on the collaborators I expect this role to be similar to the way I have either supervised students or collaborated with other Ph.d students or project partners
  • In have in one instance had to protest the methods and work of a partner (machine learning startup) in a multi-million $ project. As the only machine learning expert in the project outside of that startup, I felt that it was my responsibility. The partner vent out of business and exited the project a few months later.

(E) Proof of concept or finished product

I will build a solution to an agreed upon problem. Either this works as a collaboration where I contribute with one or more of my core competences or I take care of everything.

  • Develop and train model
  • Evaluate and visualize results
  • Deploy model


  • If the problem is relatively straight forward I will likely be able to quickly modify one of my existing solutions
  • If the data is messy or poorly balanced and if the task is complex building a solution is going to require significant development and experimentation
  • Risky projects may not result in a viable product. In this case the output of the process is going to be a report documenting the failed experiments and possible further actions and new knowledge


I have the following considerations for our Collaboration Agreement. It might not be necessary to include all of it on paper but I think it is relevant.

  • Consultant’s liability in damages cannot exceed his/her Consultant’s fee for performing the specific task, and this is irrespective of his/her being held liable for several individual claims. If the performance of the task is divided into phases, the Consultant’s maximum liability in damages will be the fee for performing the specific phase of the task.
  • Consultant provides advice and counselling to the Company but final decisions are made by the company
  • What is the consequence of a delay from Consultant's side and from the Company
  • How to handle data, how to share it and when I must delete my copy
  • I have/will acquire the hardware necessary for my own experimentation
  • Who will provide the draft of the Collaboration Agreement? I have this draft from my workers union IDA