YOLOv8 video annotation and model training app
Streamlit app for object detection with YOLOv8. Upload a model or dataset, tune hyperparameters, get real-time feedback and download the outputs.
Data Scientist
Software that learns, built to last.
Lead Consultant at ITC Infotech, with a B.Tech from NIT Tiruchirappalli and a Master’s in Data Science and AI from IISER Tirupati. I build ML models, ETL pipelines and apps, and take them from notebook to production.
training lossvalidation accuracy
Projects
A selection of my data science and machine learning work. The full source code for each project is on GitHub.
Otto Group product classification
Cleaned and analysed 70,000+ records with 93 features. A tuned LightGBM classifier reached a log loss of 0.47, within the top 30% on the Kaggle leaderboard.
Screenshot or video goes here
Car evaluation with a feedforward neural network
TensorFlow and Keras Tuner model with dropout reached 92% accuracy, up from a ~84% logistic regression baseline. Evaluated with loss curves and a confusion matrix.
Blog
Notes on Python, ML & AI engineering, MLOps and building with LLMs.
LLMs and RAG1 post
Retrieval, evaluation and building with language models.
MLOps1 post
Taking models from notebook to production.
Machine Learning1 post
Modelling, tuning and lessons from competitions.
Contact
Send a message below, or reach me directly.