Vineet Haswani Email & Phone Number

Dhule, Mh, In

Bengaluru, Karnataka, India

• Developed and implemented a comprehensive pipeline that extracts attributes from aerial imagery, such as Digital Surface Model (DSM), Solar Access Value (SAV), Pitch, and Azimuth. This innovation reduced processing time from hours to seconds and lowered costs by minimizing the need for rooftop sensors and decreasing manual effort in measuring roofs. • Developed and implemented a quality check (QC) mechanism to filter out poor-quality images and monitor real-time model accuracy. • Implemented an auto-alignment mechanism for vertical alignment along the z-axis, significantly reducing misalignment in subsequent processes. • Working on a 3D roof modeling solution using aerial imagery, aiming to reduce manual modeling time from 30 minutes to just 5 seconds. • Implemented a RAG mechanism on SQL data to extract information using prompts.

Bengaluru, Karnataka, India

• Developed an end-to-end pipeline for solving the problem of named entity recognition (NER). The pipeline includes data collection, data annotation, data preprocessing, model creation, and model deployment. • Handled SQL and No-SQL databases which were further deployed to AWS RDS and AWS Keyspaces (Cassandra).• The TensorFlow models were deployed using the TensorFlow Serving framework on a separate server.• The scikit-learn model was deployed using AWS sagemaker.• Used apache airflow for workflow management and monitoring.

Bengaluru, Karnataka, India

• Developed a deep learning model and dataset from scratch for named entity recognition (NER) problem. • The model was built using the TensorFlow framework that was capable of producing a 0.91 f1-score on our test set. • The dataset annotation was done using doccano tool (open source web application for NLP).