ANUPAMA KANDALA

Building AI-based Product Features like Rest Useful Life and various Battery Prediction use cases for Electric Vehicles using real-time telematics data and Machine Learning Regression algorithms (XGBoost, Random Forest). Engineered a scalable time series forecasting model involving Deep Learning, which is capable of handling large datasets and real-time updates. Serialised the predictive models to create a web API using Fast API, containerised using Docker and deployed it on the AWS EC2 server. Scraped Geo-location data for a Tier 1 City using Google Maps APIs and processed the data (7 million). Performed Feature Engineering, and built a Dashboard using Grafana. Generalised conclusions about population data using sample inference. Optimised the cost of the POC by 86% (saving 2.44 crores INR).

Played a pivotal role in automating the assembly process of 2-wheeler Products- Pulsar, KTM and Husqvarna, reducing manufacturing defects and increasing overall productivity by 22% with precise and reliable visual inspections using Computer Vision. Independently contributed to developing the whole Pipeline of the Proof of Concept from Image Acquisition and model Building to Model Evaluation. Built a Low-Cost Template Matching Image Processing Solution to Identify and verify the robot’s gear positioning in the machine well. Annotated the data and Developed a robust object identification model using the Detecto library for different Husqvarna bike mudguards. Built an object detection model using YOLOV5 to identify possible paint defects on KTM 2 wheeler’s petrol tanks-scratch, dust, and dent.

Developed a speech recognition model by using a convolutional neural network. Conducted hyperparameter tuning and integrated it with the image processing techniques. Tech: TensorFlow, Keras, Librosa, MFCC Techniques.

Course Work: Calculus, Linear Algebra, Probability, Pattern Recognition.