Oct 2018 – Mar 2020 · 1 yr 5 mos · Greater Seattle Area

• Applied Physics Lab: Ocean Observatories Initiative funded by the NSF. Advisor: Dr. Aaron Marburg.
• Worked on generating photo-realistic synthetic image data for augmenting object detection models where training data is not available, for an application where the model is expected to work when deployed underwater without any training data from real-setting. Evaluated the use of Neural Style Transfer to create synthetic underwater images from a raw 3D model.
• CamHD Video Analysis: Detection and Segmentation of deep-sea macro-fauna (Scaleworms) from the videos regularly collected at a hydrothermal vent located 1500 km deep in the Pacific ocean (Axial Seamount Volcano). Created a hybrid Convolutional Neural Network (U-Net + VGG16) architecture to perform instance segmentation of Scale worms with relatively very-less labeled data, which produced an Average Precision of 0.71. The challenge was to identify objects of varying size and color, sometimes camouflaged against the rich dynamic background. Such an autonomous system would be impactful in creating large scale marine population data which can be used by benthic biologists to study the diverse ecosystem in the extreme environment near hydrothermal vents, having a temperature around 400 degrees Celcius, and very high pressure. Published a research paper in WACV 2020 conference.
• Tech Stack: Python, TensorFlow, OpenCV, Deep Learning - UNet, VGG16, Neural Style Transfer.

Jul 2017 – Jul 2018 · 1 yr · Bengaluru Area, India

• Peripheral Blood Smear Analyzer (Shonit): Worked on identification and estimation of the total count of 'Platelets' and morphological classification of WBCs and RBCs (Poikilocytes). Experimented with variations of Convolutional Neural Network architectures to segment and classify the cells. Generalized the model with batchwise-random augmentations to make it robust to variations in stains and cameras. Statistically determined the uniform region of the blood-smear on which regression algorithms were applied to estimate the total count of the cells. Evaluated the performance of the models in a clinical trial study. The solution was commercially deployed at several diagnostic labs in India generating affordable diagnostic reports.
• Patents Filed:
• Hematological Parameter Computation from a Blood Smear [201841013567 (IN)].
• Performing Hierarchical Classification of Objects in Microscopic Image [201841020083 (IN)].
• Fundus (Retina) image analysis: Experimentation with Deep Learning models to identify refer-able cases of Diabetic Retinopathy.
• Developed an extensible and flexible framework on TensorFlow, to train any computational graph with real-time tracking and visualization support. Received 'Performer of the Month - Aug2017' spot award.
• Tech Stack: Python, TensorFlow, Keras, OpenCV, Deep Learning - CNN, UNet, Machine Learning, MongoDB.

Jul 2016 – Jul 2017 · 1 yr · Hyderabad Area, India

• Software Developer in the Arbitrage Trading System team (Front Office - Quant).
• Worked on the development of a robust trading system which supports heterogeneous instruments to be traded in various security exchanges.
• Quant application registration and monitoring API: Developed an API which helps to launch quant applications on the proprietary trading system with real-time process monitoring while ensuring a safe exit. This project involved concepts like daemonizing with UNIX double forking and signal handlers in Python. All the applications on the trading system were migrated to use this launcher.
• Market data loading engine: Developed efficient market-data loading and organizing software used by the proprietary trading system to trade across various security exchanges in the world. The project involved data modeling, software design, development, and integration. This system was adopted by all the trading strategy teams.
• Tech Stack: Python, Java, Perl, SQL.

Aug 2015 – Nov 2015 · 3 mos · Bengaluru Area, India

• Lab: Topic Synthesis and Analysis, Dept. of Computer Science and Automation (CSA).
• Mentors:
• Prof. M Narasimha Murty (Dean, Engineering IISC, and Professor at Dept. of CSA, IISC)
• Prof. V S Ananthanarayana (HOD, Dept. of Information Technology, NITK)
• Mr. Govind Sharma, Ph.D. in Topic Analysis, Dept. of CSA, IISC.
• Worked on Extracting and Quantifying knowledge present in research documents by a Structural organization of the text and Semantic Modelling of Natural Language presented by various authors in the research documents. Created a "CNESSI: Citation Network with Enhanced Structural and Semantic Information" dataset aimed at qualitative improvement of Digital Libraries and Indexing tools like Google Scholar, Scopus etc. Open-sourced the created dataset.
• Tech Stack: Python, NLP, Topic Modeling with Latent Dirichlet Allocation (LDA), Machine Learning - SVM.

Jun 2015 – Jul 2015 · 1 mo · Hyderabad Area, India

• Worked on Software Development at IT-Front Office, Quant department.
• Developed a Statistics Framework in Java for collecting and monitoring statistics of variables used in quant applications across the trading system. Received full-time placement offer as a result of successful completion of the project.
• Tech Stack: Python, Java, Perl, SQL.

Dec 2014 – Dec 2014 · 0 mo · National Institute of Technology Karnataka, Surathkal

• Track: Neural Networks for Data Analysis.
• Mentors:
• Prof. Bhiksha Raj, Language Technologies Institute, Carnegie Mellon University (CMU).
• Prof. Rita Singh, Language Technologies Institute, Carnegie Mellon University (CMU).
• Prof. Pulkit Aggrawal, Massachusetts Institute of Technology (MIT).
• Worked on 'Identifying Diabetic Patients with High Risk of Readmission' to possibly prevent readmissions through extra initial diagnosis. Estimated to save $59 million for a test-set of 23 thousand patients. Published in ArXiv.
• Tech Stack: Python, Association Rule Mining, Machine Learning, Feature importance analysis using ablation study.

May 2014 – Jul 2014 · 2 mos · Bengaluru Area, India

• Worked on a project for GOQii –a fitness wristband associated with phone application where a personal coach guides the user.
• Developed an analytics module to determine the 'Motivation' of GOQii users by analyzing features collected from the device and sentiments extracted from the user-coach conversations. This pilot project led to a contract with GOQii.
• Tech Stack: Python, Scalable Feature Extraction using Apache Spark, Sentiment Analysis.