Problem statement
Important feedback on the quality status of Philips equipment in the field comes from customer complaints, software logs, service notes, and replaced parts. Millions of records are collected worldwide. To identify improvement opportunities, data is captured, monitored, and analysed.
Philips has started using AI agents to make such data analysis more effective and efficient. A robust modular agent architecture is required to support various data analysis use cases and overcome time consuming manual processes to prevent or handles quality issues by enabling root-cause analysis and improving decision-making with respect to planning actions in e.g. product design and development, defect investigation, complaint handling, allocation of corrective actions (e.g. CAPA).
Objective
Design and implement a modular agentic architecture to support quality processes of our business units, included (but not limited to) product development and complaint handling workflows.
The E2E solution will help teams such as Product Design Quality and Post Market Surveillance to streamline and augment quality of respective workflows as well as increase effectiveness of planning actions, enhancing the quality of our equipment through feedback loops to our R&D engineers towards design control implementation, ultimately improving satisfaction of our customers along with benefits in case of audits by regulatory bodies (e.g. FDA).
Scope
Quality processes of Philips business units (Image Guided Therapy Systems, Precision Diagnosis, Personal Health Devices, …)
Deliverables
- Working prototype of multi-agent solution to improve processes such as product development, complaint handling and trending.
- Core building blocks of agentic architecture as reusable components (e.g. supervisor agent, RAG agents, classifier tools, agent memory, human-to-agent interactions within the agentic workflow to enable human feedback)
- Support creation of best practices for our team towards the proper use of new technologies (e.g. AI playbooks)
Skillset
- Solid Python skills (pandas, numpy, scipy, sklearn, matplotlib) and ideally some knowledge of software engineering practice
- Good understanding of MLOps practices (git workflow, unit test, CI/CD pipelines), cloud technologies (Azure, Databricks), databases (SQL, Spark, Neo4J)
- Experience with data science project steps from EDA to model evaluation and deployment
- Autonomously conduct data exploration, processing, and analytics towards creating business insights
- Preferred hands-on experience with text processing (NLP) and/or Generative AI technologies (e.g. OpenAI, LangChain, LangGraph, embedding models, vector stores)
- Highest motivation and willingness to learn new concepts and technologies, eager to make the difference and support our business units with effective AI solutions.
Mentor
Senior data scientist from the Philips Data Science & AI solutions team
Contact
See our procedure for Master Projects
Boudewijn van Dongen, b.f.v.dongen@tue.nl