Data Scientist

The Role:
We are looking for a curious and passionate Data Scientist to join our high-impact team. You’ll work directly on AI models that analyse cardiac data and further save life. This role offers the unique opportunity to see your work make a tangible difference in patient outcomes while building state-of-the-art models.

What You’ll Do:

1. Design and Implement AI Models: Develop, train, and evaluate machine learning and deep learning models for cardiac datasets and associated metadata.
   
2. Data Pipeline Development: Work with large, complex, and sometimes messy clinical datasets. Contribute to building robust and scalable data pipelines for data ingestion, cleaning, feature engineering, and labeling.
3. Model Deployment and MLOps: Deploy models to production environments and monitor their performance in real-world clinical settings. Build and maintain REST APIs for model inference using frameworks like FastAPI or Flask. Design scalable API endpoints with proper request validation, error handling, and authentication.Implement and maintain robust MLOps practices, including version control, continuous integration/continuous deployment (CI/CD), and model monitoring in a production environment (e.g., cloud platforms like AWS, Azure, or GCP).
4. Performance Optimization: Optimize model performance for inference speed and resource efficiency, crucial for deployment on various platforms (cloud, edge devices).
5. Collaboration: Work collaboratively with software engineers, data scientists, and clinical domain experts to translate clinical needs into technical requirements and deliver high-impact solutions.
6. Documentation and Research: Maintain detailed documentation of models, code, and experiments. Stay current with the latest research in deep learning, medical image analysis, and time-series analysis.
7. Regulatory and Compliance: Develop documentation in order to comply with regulatory requirements such as CDSCO, FDA etc.

What We’re Looking For:
Required:

1. Experience: 2+ years of professional experience as a Data Scientist, or a related role.
2. Education: MS/M.Tech preferred; strong BTech with relevant experience considered.
3. Programming: Proficiency in Python and experience with core data science libraries (NumPy, pandas, scikit-learn).
4. Deep Learning Frameworks: Hands-on experience with at least major deep learning framework (PyTorch).
5. MLOps Basics: Familiarity with MLOps principles, including containerization (Docker) and cloud service experience (AWS, Azure, or GCP).
6. Understanding of model evaluation metrics, cross-validation, and debugging ML systems.
7. Ability to read and implement research papers.

Preferred Skills

1. Experience with healthcare data.
2. Understanding of statistical methods and experimental design for model validation.
3. Experience with structured training pipelines such as with Pytorch Lightning.
4. Knowledge of regulatory requirements for medical devices (FDA, CE marking).
5. Experience with cloud platforms (AWS, GCP, Azure) and serverless deployments.
6. Publications in top medical journals/conferences such as ICLR, Neurips, JAMA Cardiology, EHJ, MICCAI etc.

Non-negotiable skills

1. Python + PyTorch, hands-on model building. Must be able to independently design, train, debug, and evaluate deep networks — not just call .fit(). Core of the job.
2. Deep learning. Direct experience with data (1D/2D CNNs, RNNs/temporal transformers, or similar).
   
3. Knowledge of ML evaluation on messy clinical data. Validation with clinical data, handling class imbalance, calibration, sensitivity/specificity/AUC & debugging of ML systems.
4. Ability to read and reproduce research papers. must turn a paper into working code.

Nice-to-have skills

1. Healthcare/ECG domain exposure: understands the clinical context, artifacts, and what a cardiologist actually needs (accelerates ramp-up, but teachable).
2. MLOps & deployment: Docker, a cloud platform, FastAPI/Flask inference APIs. Valuable, but a modeling hire can learn later..
3. Regulatory awareness + publications: familiarity with CDSCO/FDA/CE documentation, and papers in venues like MICCAI/NeurIPS/JACC/EHJ. A strong signal, not a gate.