Aniruddha Das Email & Phone Number

United States

Redmond, Washington, Us

Part of Substrate Intelligence: Semantic Understanding, Representations and Generation for Enterprises (SURGE). Working to apply NLP in enterprise scenarios.

Redmond, Washington, Us

Part of Substrate Intelligence: Semantic Understanding, Representations and Generation for Enterprises (SURGE). Working to apply NLP in enterprise scenarios.

Redmond, Washington, Us

Part of MSAI - Substrate Intelligence Horizontal Sciences team.Worked to gather insights from raw text data and leverage gathered insights to maximize value created for end customers.

Atlanta, Georgia , Us

Assisted ~300 students with coursework and guided them to a better grasp of various concepts in Machine Learning. Assisted with grading of assignments and proctoring of exams. Also provided one-to-many tutorials through scheduled office hours and helped debug errors arising due to the usage of multiple Machine Learning frameworks.

Atlanta, Georgia , Us

Used genetic programming and a framework that alters the development of algorithms through an automated method that starts with the best human algorithms and then dispassionately develops hybrid algorithms that outperform existing methods.Used multi-objective optimization to provide a set of optimal algorithms that span desired objectives. A human can then select the algorithm that best meets their needs. Spring 2019Led a team that developed a new generic framework that uses CGP to solve problems including but not limited to neural architecture search. The framework was developed with the aim of allowing injected domain knowledge by modularizing components in the computational pipeline. The optimal combination of these components would be searched for using Cartesian Genetic Programming before ultimately concatenating pre-processing, neural network architecture and post-processing modules to find an optimal overall solution.Fall 2018Used Cartesian Genetic Programming to minimize human participation in the discovery of neural network architectures for image classification problems by evolving from trivial initial models to reach high accuracies comparable to established convolutional neural network architectures. Leveraged Google Cloud Platform's Virtual Machine and TPU support to run evolutionary process. Spring 2018Worked on taking in a speech signal and using Bayesian Changepoint Detection (BCD) to identify points in the signal where a significant change is likely to have occurred. These found changepoints were used to trim the size of the signal by eliminating regions where the effective energy of the signal was negligible and using this portion of the signal would give us no additional information about the speech signal. Additionally, used BCD to take in a signal and assign feature data to it which could be used by Machine Learning algorithms to associate the passed in speech signal with a word.

Atlanta, Georgia , Us

Worked on time series modelling using a variation of Continuous Time Hidden Markov Models on multimodal data to construct interpretable time-to-event predictions of lapse to alcohol or drug use.

Cambridge, Ma, Us

Worked on a hot water heater project and engineered a predictive deep learning model that would be conditioned on the historical demand data of a specific household. Learned the households hot water demand (collected every minute) and predicted future demand for the next 24 hours using the predictive deep learning model. The model-generated predictions were used to schedule heating times and prevent growth of harmful Legionella bacteria. Evaluated existing research, developed and tested multiple models and approaches to find one that worked and quantified the potential impact of the research. Work is published in the Elsevier - Engineering Applications of Artificial Intelligence journal.Also worked on a 6D pose estimation problem where mesh models of objects were collected along with 3D point cloud data of different environments and fed to a novel engine to generate data. The data was collected by interfacing with the iPad Structure Sensor and Xbox Kinect Sensor. This collected data was then used to train a deep learning model to estimate the 6D pose of a variety of objects in different environments

Was a part of the time series feature and anomaly detection team that worked towards early identification of anomalies in different contexts. Devised a method to get a OneClassSVM to work with time series data and identify novelties in it by taking contextual information into account. Additionally gave normalised probability scores for the anomalous regions. Additionally, explored real-time learning methods to detect anomalies in videos by identifying the norm and then classifying anomalies as video sequences that violate this norm.

Bengaluru, Karnataka, In

I created an endpoint on an API that ran on Google App Engine Standard. This API could be queried to get information about states in a workflow. Wrote Java code which received information from the backend and extracted relevant information to return the names of the states, together with information on their transitions i.e. conditions under which the transitions take place and the states they transition to. The goal was to use this information to visualize the workflow while keeping the code as generic as possible. The returned information was used by a client side program written by me to visualize the workflow. I used D3.js to visualize the workflow using the information returned from querying endpoint. Additional information about these transitions was returned such as the sentiment of the condition under which the transition occurs. i.e. if a transition occurs to a terminal state when the application is rejected, the sentiment associated with this transition is negative. As the code had to be generic, and sentiment values could not be hard coded as the platform scaled, I queried Google’s Natural Language API to find associated sentiments. Also implemented a caching mechanism to minimize expensive API calls.

Master Of Science - Ms, Machine Learning

Bachelor Of Science - Bs, Computer Science

Icse, Isc, Science