Senior AI Engineer

• Apr 2020 - Feb 2021

Research And Development Engineer

• Jul 2018 - Mar 2020

Title: Blockchain platform design and deployment: • Part of the team that designed and Implemented Hyperledger Fabric based Blockchain platform for Honeywell. • The platform serves to the backend of the world’s first aircraft part trading platform: https://www.godirecttrade.com/ • Implemented KAFKA based ordering service for managing transaction commits, propagation to nodes, avoiding double spend, etc. • Designed and implemented role based access control for the Blockchain… Show more Title: Blockchain platform design and deployment: • Part of the team that designed and Implemented Hyperledger Fabric based Blockchain platform for Honeywell. • The platform serves to the backend of the world’s first aircraft part trading platform: https://www.godirecttrade.com/ • Implemented KAFKA based ordering service for managing transaction commits, propagation to nodes, avoiding double spend, etc. • Designed and implemented role based access control for the Blockchain platform. • Docker swarm based communication between nodes in the Blockchain network. Title: AI powered Aircraft part document reading: • Deep learning based methods for text area identification and content detection in scanned aircraft documents. Title: Market Research for Adjacent Market Analysis: • In-house sensor market study in railway applications. • Radar and Telemetry market study to identify potential gaps and applications. • Potential Blockchain applications in aerospace. Show less Title: Blockchain platform design and deployment: • Part of the team that designed and Implemented Hyperledger Fabric based Blockchain platform for Honeywell. • The platform serves to the backend of the world’s first aircraft part trading platform: https://www.godirecttrade.com/ • Implemented KAFKA based ordering service for managing transaction commits, propagation to nodes, avoiding double spend, etc. • Designed and implemented role based access control for the Blockchain… Show more Title: Blockchain platform design and deployment: • Part of the team that designed and Implemented Hyperledger Fabric based Blockchain platform for Honeywell. • The platform serves to the backend of the world’s first aircraft part trading platform: https://www.godirecttrade.com/ • Implemented KAFKA based ordering service for managing transaction commits, propagation to nodes, avoiding double spend, etc. • Designed and implemented role based access control for the Blockchain platform. • Docker swarm based communication between nodes in the Blockchain network. Title: AI powered Aircraft part document reading: • Deep learning based methods for text area identification and content detection in scanned aircraft documents. Title: Market Research for Adjacent Market Analysis: • In-house sensor market study in railway applications. • Radar and Telemetry market study to identify potential gaps and applications. • Potential Blockchain applications in aerospace. Show less

Postgraduate Student

• Jul 2015 - Jul 2018

Title: Analysis of homogeneity and heterogeneity of breast cancer sub-types using Deep Learning: • Modelled the genomic molecular profile data using Semi Supervised Non-negative Matrix Factorisation (SSNMF) to predict the predominant breast cancer subtypes. • Classified the data as homogeneous and heterogeneous based on the entropy function of classifier. • Developed residual neural network model to predict the multiple subtype present in whole slide H&E stained image of breast cancer… Show more Title: Analysis of homogeneity and heterogeneity of breast cancer sub-types using Deep Learning: • Modelled the genomic molecular profile data using Semi Supervised Non-negative Matrix Factorisation (SSNMF) to predict the predominant breast cancer subtypes. • Classified the data as homogeneous and heterogeneous based on the entropy function of classifier. • Developed residual neural network model to predict the multiple subtype present in whole slide H&E stained image of breast cancer tissue. • Generalised the model to work across various datasets by doing colour normalisation. Course Projects: 1. Deep CNN for Facial Expression Recognition • Implemented a model in python using Tensorflow, which classify the 7 different facial expressions. • Architecture of the model has parallel feature extraction block, which was inspired by GoogleNet. • Compared the results with 3 layer LeNet architecture. 2. Neural network to predict income exceeding a specific value • Implemented neural network from scratch in python using numpy with variable number of neurons. • Trained neural network through back propagation with momentum gradient descent. 3. Image search to extract features and retrieve similar images from database using histogram refinement • Implemented GUI enabled Content based image retrieval, using histogram refinement, in MATLAB. • Compared the results with another method based on colour, shape and texture. 4. Determining Optical Flow • Determined optical flow from set of images having camera ego motion in MATLAB. • Implemented using the brightness constancy assumption as well as feature-based method. Show less Title: Analysis of homogeneity and heterogeneity of breast cancer sub-types using Deep Learning: • Modelled the genomic molecular profile data using Semi Supervised Non-negative Matrix Factorisation (SSNMF) to predict the predominant breast cancer subtypes. • Classified the data as homogeneous and heterogeneous based on the entropy function of classifier. • Developed residual neural network model to predict the multiple subtype present in whole slide H&E stained image of breast cancer… Show more Title: Analysis of homogeneity and heterogeneity of breast cancer sub-types using Deep Learning: • Modelled the genomic molecular profile data using Semi Supervised Non-negative Matrix Factorisation (SSNMF) to predict the predominant breast cancer subtypes. • Classified the data as homogeneous and heterogeneous based on the entropy function of classifier. • Developed residual neural network model to predict the multiple subtype present in whole slide H&E stained image of breast cancer tissue. • Generalised the model to work across various datasets by doing colour normalisation. Course Projects: 1. Deep CNN for Facial Expression Recognition • Implemented a model in python using Tensorflow, which classify the 7 different facial expressions. • Architecture of the model has parallel feature extraction block, which was inspired by GoogleNet. • Compared the results with 3 layer LeNet architecture. 2. Neural network to predict income exceeding a specific value • Implemented neural network from scratch in python using numpy with variable number of neurons. • Trained neural network through back propagation with momentum gradient descent. 3. Image search to extract features and retrieve similar images from database using histogram refinement • Implemented GUI enabled Content based image retrieval, using histogram refinement, in MATLAB. • Compared the results with another method based on colour, shape and texture. 4. Determining Optical Flow • Determined optical flow from set of images having camera ego motion in MATLAB. • Implemented using the brightness constancy assumption as well as feature-based method. Show less

Associate Software Engineer

• Jul 2014 - Jun 2015

Graduate Student

• Jul 2013 - Jun 2014

Title: Automatic Speaker Recognition System: • Mel Frequency Cepstrum Coefficient (MFCC) of each individual is extracted as the features. • Dynamic Time Warping(DTW) is used to compare with test sample to identify the individual. Title: Design of Floating point Square Root Operation for ASIC and its implementation on FPGA: • Designed two iterative non restoring algorithms using IEEE754 single-precision floating-point format. • Implemented algorithms in verilog and compared… Show more Title: Automatic Speaker Recognition System: • Mel Frequency Cepstrum Coefficient (MFCC) of each individual is extracted as the features. • Dynamic Time Warping(DTW) is used to compare with test sample to identify the individual. Title: Design of Floating point Square Root Operation for ASIC and its implementation on FPGA: • Designed two iterative non restoring algorithms using IEEE754 single-precision floating-point format. • Implemented algorithms in verilog and compared their performance and power efficiency in FPGA. Show less Title: Automatic Speaker Recognition System: • Mel Frequency Cepstrum Coefficient (MFCC) of each individual is extracted as the features. • Dynamic Time Warping(DTW) is used to compare with test sample to identify the individual. Title: Design of Floating point Square Root Operation for ASIC and its implementation on FPGA: • Designed two iterative non restoring algorithms using IEEE754 single-precision floating-point format. • Implemented algorithms in verilog and compared… Show more Title: Automatic Speaker Recognition System: • Mel Frequency Cepstrum Coefficient (MFCC) of each individual is extracted as the features. • Dynamic Time Warping(DTW) is used to compare with test sample to identify the individual. Title: Design of Floating point Square Root Operation for ASIC and its implementation on FPGA: • Designed two iterative non restoring algorithms using IEEE754 single-precision floating-point format. • Implemented algorithms in verilog and compared their performance and power efficiency in FPGA. Show less