What Kinds of Projects Do S2DS Participants Work On?

Updated

S2DS participants work on real commercial projects for client organisations over the five-week programme, not on simulated exercises. Working in small teams with agile management methods and the mentorship of an experienced data scientist, they produce deliverables with real impact for those clients. The projects span applied areas such as large language models, generative AI, machine learning and natural language processing, and each one is built on the analytical skills participants bring from their PhD or MSc.

How the project work is structured

From day one, participants are immersed in the day-to-day life of a data scientist. That means:

  • Client-facing teamwork — you work in a small team directly with a client, rather than on a solo assignment.
  • Agile management — teams use agile methodologies to plan and deliver within the five-week window.
  • Mentorship — an experienced data scientist guides the team technically and professionally.
  • A concrete deliverable — the programme's stated aim is a result with real impact for the client, not just a learning exercise.

This structure is what distinguishes S2DS from purely classroom-based training: the commercial context is the curriculum.

Examples of past S2DS projects

S2DS publishes past projects that show the range of work. These illustrate the kind of brief a team might receive:

Project area What the team did
Large Language Models (LLMs) Built a proof-of-concept to automate refactoring notebooks into production-ready scripts using LLMs; established a framework for LLM-based automation and flagged challenges for the partner to build on
Generative AI Used generative AI tools and a fine-tuned MLflow pipeline to predict gender in a journalism corpus
Machine Learning Built an ML model using eddy current measurements to predict the depth of flaws in heat exchanger tubes, improving maintenance regimes for electricity generating equipment and reducing unnecessary downtime and power losses
Natural Language Processing (NLP) Improved classification of food hazard alerts using NLP, developing an ensemble pipeline that combined rule-based matching and transformers with classifiers

Two things stand out across these examples. First, the problems are operational — maintenance scheduling, content classification, code production — rather than toy datasets. Second, the outputs are framed as things a partner can take forward: a framework, a pipeline, a proof of concept.

What this means if you are deciding whether to apply

The project model suits you if you want evidence of applied, client-facing data science work rather than another certificate. Because teams deliver for an actual client under time pressure, the experience is designed to mirror a data science job — one Spring 2025 alum, Athul Sivan, described the balance between learning and working as well handled, with the boot camp simulating a data science role while teaching technical and interpersonal skills.

It is less suited to someone seeking a deep, single-topic technical course; the value lies in breadth of commercial exposure, teamwork and business acumen alongside the technical delivery.

Where to look next

The past projects page on s2ds.org is the best place to see the current spread of client briefs, and the programme also runs a podcast, Science to Data Science and Beyond!, featuring interviews with alumni about their career journeys. If you want to judge fit, compare the project examples above against the kind of work you want to be doing in your first industry role.