Production Planning Optimization in the Semiconductor Industry


The internship can take place in Paris or Montpellier; remote work is possible if imposed by regulation. Compensation according to regulation. This internship could lead to a permanent position in Operations Research at DecisionBrain, working on various projects in the fields of Manufacturing, Workforce Management, Logistics and Maintenance. The start of the internship is flexible but is expected to be in Spring 2024


DecisionBrain has developed a tactical production planning software for a client in semiconductor manufacturing, that this internship aims to improve. Classical production planning problems aim at satisfying the demand on a discrete planning horizon, under resource capacity constraints with the objective of reducing the total cost (typically lost sales costs, tardiness costs and inventory costs). Semiconductor manufacturing is the industry that processes integrated circuits (a.k.a. chips). In our case study, the production process is focused on the testing of the chips which can take up to 40 steps and has an average delay (cycle time) of 15 days to process all the steps.
The client production process is challenging with large datasets and several complex feature such as:
• Multi-level production,
• Choice between various lead times (effective delay to complete a step) for some steps,
• Alternatives of product consumed, alternatives of resources used,
• And more…

In current scenarios, a few thousand demands must be satisfied, with hundreds of resources (machines and workforce) and hundreds of products. In the near future, the number of demands will increase as the business of the client grows. As the optimization problem is already quite large, the exact optimization model performs poorly when the size of the instances is too large. Hence, the aim of the internship is to design and implement metaheuristics and decomposition procedures to complement the current exact approach and to improve performance.

The internship consequently consists in:
• Performing a state of the art on production planning decomposition methods,
• Proposing and implementing metaheuristics and decomposition algorithms that rely on the implemented
exact method,
• Comparing the performance of the different algorithms.
The implementation of the solution approaches should be done in Java.


Candidates must be M2 level students (2nd year of MSc or last year of “cycle ing ́enieur”). They must have a solid background in computer science, good programming skills (preferably in Java), and a particular liking for Operations Research.

Candidates must send their CV, a letter of motivation and their grades to [email protected].