Abhinav Singh

Experience

• 

  Tata Consultancy Services

  Mar 2012 - Dec 2015

  Systems Engineer

  Worked as a Backend Java Web Developer and designed and developed new web flows and monitored enhancement and maintenance of the existing web application portal as per client's requirements for Citibank Inc. USA.Led the web application code upgrade team to deliver the application code for development before every release on time.Automated the process of upgrading the enterprise web application code with new framework jars and configuration files provided by the framework team by developing script-based software using Java and ANT.Worked as a software automation tester for the client's internal application. Show less

• 

  EDIFIXIO INDIA PRIVATE LIMITED

  Dec 2015 - Sept 2016

  Software Engineer

  Worked on Information Manager of Oracle Knowledge to create a workflow for content authoring, editing, review, and publishing environment to support the entire knowledge lifecycle of Schneider Electric, FranceWeb Development of Schneider Electric Website FAQ Section

• 

  Wipro

  Apr 2018 - Mar 2020

  Senior Software Engineer

  Developed a highly optimised Optical Character Recognition (OCR) pipeline for extracting records from patients’ prescription for an insurance company using computer vision(Maximally Stable Extremal Regions (MSERs)) and deep learning advances such as bi-directional Long Short Term Memory (LSTMs), Connectionist Temporal Classification (CTC) and convolutional neural nets (CNNs).Developed a Conveyor Belt Damage Detection system using computer vision and Mask-RCNN to tackle damage caused by foreign objects (iron rods and specs) in coal distribution system and achieved an accuracy of 84% (test recall of 86%).Aircraft Image Inspection: Designed and developed lightning strike damage detection and quantification machine learning model using RetinaNet (with ResNet101 backbone) neural architecture. Successfully handled the challenges of a smaller number of useable images, of varying resolution and from different sources, using computer vision techniques and data augmentation to achieve an outstanding test recall of 88.06% (Target Range: 60% - 80%).Medical Image Solutions: Chest X rays segmentation using customised UNet neural architecture (Accuracy - 97.95% and Intersection over Union- 0.915), classification of tuberculosis manifested or normal chest x-ray using VGG19 neural network (accuracy – 91.67% after series of image processing techniques and data augmentations) and automated label generation using deep neural network (CNN for feature generation and RNN for Image label captioning) for a reputed customer in Healthcare Sector. White Paper - https://builders.intel.com/docs/aibuilders/wipro-medical-imaging-solutions-powered-by-intel-xeon-processors.pdfDisastrous Event Classifier: Disastrous Event Analyser using Natural Language Processing to classify an untoward report of incident from different external sources of information and generated a relationship graph to analyse and correlate information from different sources. Achieved a test accuracy of about 87.67%. Show less

• 

  Ericsson

  Mar 2020 - Jul 2021

  Data Scientist

  Worked on a system log messages’ categorization model to detect anomalous log events using natural language processing (NLP) and clustering techniques which would help in predicting network-based CIP Diameter rejections.Smart Synthetic Sequential Data Generation – Developed a POC to generate network based time series data with high fidelity and secure data using a Generative Adversarial Network (GAN) based DoppelGANger network to deal with data scarcity and data privacy.User Behaviour Analytics (UBA): Autonomous Security Threat Detection – Developed a system to analyse command level, session level and user level activities to detect anomalous behaviours using NLP for text processing, Isolation Forest for modelling and Statistical Inferencing (Modified Z-score) method to find deviations on daily aggregated user behaviour along with Ranking of users based on number of anomalous days across different categories. Intelligent Ticket Solution Prediction System – Developed an ”Intelligent Ticket Solution Problem Prediction System” based on Machine Learning to help provide the possible resolution of customer issue as soon as ticket is received by the Front office team based on historical data using LDA, index based neural network, Latent Sequence Model and One Shot Encoding Neural network for generating best possible solution. The solution decreased the time to resolution by 75%. Show less

• 

  JPMorgan Chase & Co.

  Aug 2021 - Mar 2022

  Senior Associate
• 

  Hexagon Capability Center India

  Mar 2022 - now

  Technical Lead