Project overview:

  • Goal: To be able to translate technical reports (pentest, vulnerability assessment, red teaming) into various languages in order to enable business in other (mainly European) countries.
  • Location: Amsterdam / Eindhoven
  • Timeframe: 3 months
  • Starting: TBA
  • Complexity: Medium
  • Team: CTO
  • Supervisor: Ralph Moonen

As a student, you have:

Education:

An HBO or WO level of education in the relevant domain.

Technical skills:

  • AI
  • LLM
  • Cloud


Soft skills:

  • The ability to work well in an international team environment;
  • Good communication skills
  • Organized.
  • Clear documentation writing skills.


The project you will be working on:

Secura wishes to investigate the feasibility and practicality of translating their technical reports to other languages, using LLM’s. Many customers want reports in their native language, while Secura’s team can only support Dutch and English. In order to be able to open up many more European countries, it would make sense to use LLM’s to translate reports. However, the use of available commercial services such as ChatGPT is not possible because the contents of our reports is too sensitive and customers forbid the processing of those reports by 3rd parties.

For this reason, we wish to have the question answered if it is possible to use LLM’s such as OpenLLM or BLOOM https://huggingface.co/bigscience/bloom#model-details to build an on-premise or at least self-hosted version that can translate technical (pentest) reports accurately. To start off we should focus on European languages, mainly Dutch, English, German, French, Spanish, Norwegian, Finnish, Danish, Italian and Swedish. If time permits, other languages as included in the BLOOM model such as Arabic, Greek or Chinese might be included.


As an intern, you will be responsible for the following:

  • Gathering requirements;
  • Generate shortlist of LLM’s that could do the job;
  • Select LLM together with the CTO and IT;
  • Implement LLM in self-managed environment (together with CTO and IT);
  • Test, train and optimize LLM for this specific use-case;
  • Validate results by manual evaluation with native speakers;
  • Document the process.


What our selection process looks like

Selection process in steps

Meet our recruitment team

Eefke Ruisbroek

Recruiter

Sara Busscher

Recruiter

Stan de Jong

Recruiter