In the summer of 2023, I had the opportunity to intern as a Data Science Intern at Philips, Manipal. This experience marked one of the most exciting phases of my academic and professional journey, where I worked on applying machine learning to real-world healthcare data---a domain where every improvement has the potential to make a tangible difference in people's lives.
Tackling Real Healthcare Data Challenges
My project revolved around designing and optimizing a binary classification pipeline. The dataset combined survey data and equipment records, which I cleaned, merged, and transformed into a unified training set. This improved data usability by almost 30%, enabling more reliable model development.
One of the biggest challenges I encountered was class imbalance---a common issue in healthcare datasets. For example, positive cases were far fewer than negative ones, which could bias predictions. To address this, I applied resampling techniques such as SMOTE, which boosted accuracy and AUC scores by 5--8%.
Building and Improving Models
From scratch, I engineered classification models using Random Forest and XGBoost, fine-tuning them to achieve 85--90% accuracy on real-time healthcare data. Beyond performance metrics, I focused on generalization, ensuring the models could handle unseen data effectively---a crucial requirement in healthcare applications where reliability is non-negotiable.
Learning Beyond the Code
This internship was more than just writing machine learning pipelines. I learned the importance of:\ - Data fusion: integrating multiple data sources for a more holistic view.\ - Ethics in AI: ensuring fairness in healthcare applications.\ - Collaboration: working with mentors and peers to approach problems systematically.
Key Takeaways
- Real-world data is messy, but that's where the real learning happens.\
- Techniques like SMOTE are not just academic---they make a measurable difference in practice.\
- Accuracy is important, but so is interpretability and trustworthiness, especially in healthcare.
Looking Ahead
My internship at Philips helped me appreciate the power of data science in healthcare innovation. The blend of technical problem-solving and meaningful impact strengthened my resolve to pursue a career at the intersection of AI and real-world applications.
I'm deeply grateful to my mentors and peers at Philips for guiding me through this journey. This experience laid a strong foundation that continues to influence how I approach data-driven problem-solving today.