Thesis topic: Single-stage object detection using 3d point cloudsThe thesis provides an approach to reduce the model size of the existing state-of-the-art 3D object detection networks to enable execution on mobile/edge devices. Despite compression, the models retain maximum performance and real-time capability of the original model. Instead of the conventional approach like quantization/pruning, the thesis deals with modification in the layer structure of the ConvNets.- The new… Show more Thesis topic: Single-stage object detection using 3d point cloudsThe thesis provides an approach to reduce the model size of the existing state-of-the-art 3D object detection networks to enable execution on mobile/edge devices. Despite compression, the models retain maximum performance and real-time capability of the original model. Instead of the conventional approach like quantization/pruning, the thesis deals with modification in the layer structure of the ConvNets.- The new models are at least 5.5x smaller than the baselines such as PointNet and Complex YOLO- They are also up to 2x faster with little/no impact on performance- Pareto optimization proves that the new models are more optimal than the baselines Show less

My tasks are to incorporate the predictive analytics and anomaly detection models into the existing workflows, to provide end-to-end object detection solutions to clients right from capturing and annotating images of mechanical instruments and PLC devices to building, training and testing the object detection models that can detect those instruments/devices.Tasks:- modeling ML algorithms for anomaly detection on near-hardware raw signals like (brakes, conveyors, pressure sensors… Show more My tasks are to incorporate the predictive analytics and anomaly detection models into the existing workflows, to provide end-to-end object detection solutions to clients right from capturing and annotating images of mechanical instruments and PLC devices to building, training and testing the object detection models that can detect those instruments/devices.Tasks:- modeling ML algorithms for anomaly detection on near-hardware raw signals like (brakes, conveyors, pressure sensors etc.)- preprocessing the data (cleaning, transforming and performing feature engineering)- time-series forecasting using CNN-LSTM- integrate the data pipeline with message queues enabling smooth incorporation with the existing workflow- capturing and annotating images to generate a train and test set for object detection models- building, training, hyperparameter optimization and comparison of various object detection models built on YOLO, MobileNet and other real-time detectors for the generated imagesAdditionally, I also work on various optimization techniques like unsupervised feature selection and neural architecture search. Skills used: Python, Tensorflow, Keras, scikit-learn, Numpy, Pandas, matplotlib, Tensorboard, Microsoft VoTT(annotating images)Being a part of the joint research project between Resolto Informatik GmbH and Fraunhofer-IOSB, Lemgo; I was involved in the development of object detection models on 3D point clouds.Project details:https://www.its-owl.com/projects/innovation-projects/details-innovation-projects/?tx_projekte_projekte%5Baction%5D=show&tx_projekte_projekte%5Bback%5D=582&tx_projekte_projekte%5Bcontroller%5D=Project&tx_projekte_projekte%5Bproject%5D=278&cHash=290b8c52849ecbfed9b2952684be0d9f Show less

As a Software Analyst, my tasks include :- co-ordinating with the pre-sales team to pitch up business solutions to the clients.- develop scripts for migration and ingestion of data from Oracle warehouse to Hive tables - developing Spark SQL queries for the existing SQL queries on the Oracle database.- testing, analyzing and developing reports on various big data tools such as Snowflake, Paxata etc.

As a member of the big-data team, my tasks are :- develop solutions for data ingestion and migration into the Hadoop ecosystem for banking clients such as ABN-AMRO, CIBC etc.- conduct proof-of-concept(PoC) using Hive and Spark on the migrated data.- client-interaction for requirement analysis as well as stake-holder management.- pitch-up big data use-cases to the clients

As a Mainframe developer, my tasks are :- develop COBOL drivers for interaction with Java clients- create and maintain SQL databases on DB2- develop programs using COBOL for ingestion and querying DB2 tables- client interaction for requirement gathering