INTERNSHIP - Machine Learning Intern

Tech Soft 3D is the leading provider of engineering software development toolkits and industrial applications for CAD/CAE data conversion, visualization and data publishing.
We are looking for TWO INTERNS to join the HOOPS AI team and help develop and validate CAD‑ML workflows that bring tangible value to the industry. Our goal is to introduce machine‑learning models into our library in a much more accessible and pragmatic way, making them easier to use and integrate for real engineering scenarios.

Subject: Working with 3D models for Machine Learning workflows

We are offering two internship positions designed to cover complementary topics along the CAD–ML pipeline. The first internship aims to make CAD parts much easier to find and reuse by building a richer, more consistent layer of metadata and search around them. The second targets agentic workflows for CAD analysis and interpretability using the reasoning capabilities of LLMs.

Project 1: Intelligent Metadata Layer for CAD

This internship aims to build a richer, more consistent metadata layer around CAD parts so that they are easier to search, filter and reuse. Starting from existing CAD datasets, the intern will design a simple but expressive RAG pipeline using metadata schemas (e.g. name, description, material, application domain, project, supplier) and generate realistic PLM-style information when it is missing. On top of this metadata layer, the intern will build a search experience where users can find relevant parts using natural language queries or by example parts and will systematically evaluate how much this improves the discoverability and reuse of CAD data.  The outcome will be a prototype implementation inside HOOPS AI, clear API design guidelines, and notebooks demonstrating semantic search combining shape and text retrieval.

Project 2: AI Agent layer for Interactive CAD Intelligence

This internship aims to extend the HOOPS AI toolkit with an “agent layer” that connects to multiple LLM providers (OpenAI, Claude, Gemini, Mistral, etc.) and enables natural-language workflows over CAD data. Starting with key HOOPS AI components (e.g., retrieval and detailed B-Rep inspection), the intern will design a provider-agnostic agent runtime where the LLM can call HOOPS AI functions as structured tools and produce reliable, explainable results. On top of this foundation, the intern will build 2–3 end-to-end agent scenarios. Such as: “find parts similar to this one and explain why,” “inspect a body and summarize manufacturing-relevant risks,” or “prepare a dataset and run an inference pipeline” and compare response quality across different LLM backends. The outcome will be a prototype implementation inside HOOPS AI, clear API design guidelines, and notebooks demonstrating agent-driven CAD analysis.

What you’ll do: 

As an Intern, your work will focus on applying state-of-the-art algorithm techniques to solve key challenges in the CAD industry, with a focus on minimizing manual processes and maximizing efficiency.

• Conduct literature reviews on machine learning techniques for CAD-oriented applications.
• Select and implement the best Python libraries for semantic search and agentic tooling.
• Train and validate models on academic datasets and extend their application to industrial datasets.
• Explore strategies to improve generalization, efficiency, and automation
• Compile findings into comprehensive reports and propose methodological improvements.

What we’re looking for:

• You’re currently enrolled in a master’s program (BAC +5) in data science, computer science, applied mathematics, or a related field.
• You have solid knowledge of machine learning concepts and Python.
• You are motivated, resourceful, and comfortable with open-ended research challenges.
• You communicate effectively and have a good command of English.

Bonus points if you have:

• Experience using or implementing text embeddings and semantic search is a plus and would be highly valuable
• Familiarity with CAD concepts, especially Boundary Representation (BRep) formats is a plus.
• Experience in machine learning frameworks such as PyTorch, HuggingFace, PydanticAI, Haystack, LangChain, LlamaIndex, crewAI etc
  

Location: BIOT - Sophia-Antipolis 

Duration: 5-6 months, desired starting date in April

Work hard, play harder. Here at Tech Soft 3D, we're solving some of the toughest engineering problems—but without taking ourselves too seriously. Expect to collaborate with exceptionally smart, down-to-earth people in a relaxed, start-up-energy environment.

We’re Tech Soft 3D => https://www.youtube.com/watch?v=HAgB3OW1WsM