I'm Anagha - a Data Scientist and ML Engineer who likes making messy, real-world data behave. I work across NLP, computer vision, and cloud infrastructure, mostly on AWS, and I care a lot about building things that stay reliable once they leave the notebook.
AWS Certified (ML Specialty, GenAI Developer Professional, ML Associate, Cloud Practitioner) - happy to talk shop about any of it.
- ML / AI: PyTorch, TensorFlow, Scikit-learn, XGBoost, SHAP, LIME, MLflow, spaCy, NLTK, OpenCV
- Cloud: AWS (SageMaker, Lambda, S3, Bedrock, and the rest of the alphabet soup)
- Data & DevOps: Airflow, Kafka, Docker, CI/CD, REST APIs
- Languages: Python, SQL, R, Java, C++, C#
I'm especially interested in model interpretability and building ML systems that people can actually trust and audit - not just ones that score well on a leaderboard.
modelstamp - a small Python library
that adds integrity verification and dependency-drift detection to persisted ML
models (scikit-learn, joblib, pickle), so a model.pkl comes with a receipt instead
of just a hope that it still works.
PyPI · Docs · Feedback wanted
- modelstamp - artifact integrity and environment verification for persisted ML models
- aws-textract-document-analyzer - document intelligence built on AWS Textract
- Hate-Speech-Detection - comparing classic NLP classifiers on an imbalanced dataset
- Automatic-Depression-Detection - applied ML on mental-health signals
- Facebook-AdCampaign-Analysis - ad performance and campaign analytics
- Parallel-Programming-Sudoku - a Sudoku solver fast enough to finish before you've picked a Netflix show 🎬
- Building ML pipelines that survive contact with production
- Model interpretability and explainability in practice
- Getting started with open-source or small side projects
If you have questions or want to chat, don't hesitate to open an issue!
📧 Email: anaghagdhekne7@gmail.com 🔗 LinkedIn: Anagha Dhekne