Deepika Anantha Padmanaban

Arriver builds software for the delivery of ADAS and AD solutions, focusing on sensor perception.

• Worked on experimental prototype for adapting proprietary object detection model to work with latest requirements from customers for Traffic Sign Detection using different network structures and hyper-parameters and analyzing detection results.
• Developed features of KPI report building framework for analyzing the performance of object detection model.
• Hands-on experience with Test Driven Development. Worked with peers on writing system level tests, component tests and feature requirements. Performed Fault Tree Analysis to assess risks and full coverage of unit tests for every functionality.
• Implemented features for proprietary visualization tool that uses database of detected objects for identifying errors in annotations. Developed code for tracking and visualizing detected object
• Tested the developed features using existing HiL and SiL environments.
• Involved in practices such as pair-programming, code review and team building activities like hackathons.
• Helped other team members get started with Machine Learning and voluntarily involved myself in various tasks involved in ADAS perception development to broaden my knowledge.

Tools: TensorFlow, Python, C#, C++, React, JS

Zenseact AB is an Autonomous Driving research company that I collaborated with for my Master Thesis.

Thesis Topic: Identification of fundamental driving scenarios using Machine Learning.

• Selected appropriate data representing driving scenarios and created data pipeline for ML algorithm
• Designed two-staged end-to-end unsupervised ML model to detect driving scenarios from time-series data from CAN sensors. Developed stacked sparse autoencoder model for feature extraction over sliding windows of time-series data and clustered the extracted features using k-Means algorithm.
• Compared my unsupervised techniques against state-of-art rule based scenario identification and an existing Bayesian Non-parametric approach. My model could identify all base scenarios expected from the data, similar to the existing rule-based technique unlike the Bayesian approach.

Tools: PyTorch, Python, Pandas, NumPy, Scikit-Learn

Athenahealth develops cloud-based healthcare products and services, where I worked with the test automation team.

• Automated test suite for two critical products, athenaCollector and Patient Portal, using Ruby-Watir web driver.
• Developed custom framework for UI automation tests of products.
• Maintained the automation test suite and ran them to ensure smooth releases of products.
• Developed scripts to schedule runs and reporting e-patches.
• Analyzed root causes of issues reported and assigned them to the respective teams.

Tools: Ruby, SQL

Intern in the Low-level platform team that worked with connectivity testing in Amazon devices.

• Automated testing of Wi-Fi connectivity in Amazon devices using Python.
• Involved in manual Bluetooth and Wi-Fi connectivity testing on Amazon devices.
• Assisted my mentor in protocol testing of Bluetooth connectivity using Packet Sniffers.