Master of Science (Geodesy and Geoinformation Science)

2016 - 2020

Technische Universitat Berlin, Germany

Specialization: Computer Vision and Machine Learning

Bachelor of Technology (Electrical Engineering)

2008 - 2012

West Bengal University of Technology, Kolkata, India

National Conference on Control Communication and Device Electronics, VOL 3, July

2014

Concept towards Traditional Water Supply System using Automation

ULAB Journal of Science and Engineering, Vol 4, November

2013

Prosthetic Arm: Its digital realization

International Journal of Advance Research in Electrical, Electronics and Instrumentation Engineering, Vol 2, Issue 10, October

2013

Study on Cardiovascular Disease and Compression of Generalized ECG Signal to Support Biomedical Advancement

International Conference on Information Processing, Springer, Vol 292, August

2012

Study of Sampled Data Control Approach towards Prosthetic Eye Model for Initiating the Non-Deterministic Aspect

Free Space Detection for Autonomous Vehicle using High Frequency Radar Sensor

October 2018 - March 2019

• Conducted noise reduction on high-frequency radar sensor data and transformed coordinates from the vehicle's coordinate system to the global coordinate system.
• Implemented kNN, DBSCAN, and grid-based DBSCAN algorithms to detect occupancy.
• Integrated the prototype with the provided sensor fusion framework.

C++ , Visual Studio, Cloud Compare, Cassandra, LaTeX

Computer Aided Induction Motor Design

September 2011 - April 2012

• Investigated foundational principles of induction motor design through literature review and analysis.
• Developed a generalized design framework to accommodate various motor specifications and requirements.
• Conducted comprehensive performance analysis to optimize efficiency, torque, and power output.
• Collaborated with industry experts to validate design methodology and refine implementation for real-world applications.

C++, Matlab, LaTeX, AutoCAD

Building Outline Extraction using Convolutional Neural Networks and Their Trasferability to Unseen Areas

August 2019 - January 2020

• Curated a diverse dataset of 3-band monocular high-resolution satellite images for analysis.
• Implemented state-of-the-art convolutional neural network (CNN) architectures, including FCN, SegNet, U-Net, Deep Residual U-Net, PSPNet, and Mask RCNN for building footprint extraction.
• Applied various data augmentation techniques such as random rotation, clipped zoom, flipping, luminance adjustment, histogram equalization, blur, and noise addition to enhance model generalization.
• Conducted rigorous performance evaluation of the trained models on different satellite image datasets to assess their accuracy and robustness.
• Investigated the transferability of the trained networks across different geographic regions.
• Achieved 72% IoU and 97% accuracy from U-Net and 85% IoU and 94% accuracy from Mask RCNN.
Github Link

Python, Keras, Tensorflow, OpenCV, Deep Learning, Jupyter Notebook, Google Colab, LaTeX

Data Scientist

December, 2020 - Present

Teraki GmbH, Berlin, Germany

• Contributed to the research and development of Autonomous Driving stack in both constrained and unconstrained environments, specifically targeting the Robotics and Logistics industry. Primarily focused on enhancing Map Generation, Localization, Navigation, and Automation capabilities.
• Implemented a generalized point cloud registration algorithm applicable to both structured and unstructured point cloud data.
• Engineered a non-linear unconstrained optimization algorithm specifically tailored for point cloud registration tasks targeting precise map generation.
• Integrated Lidar, IMU, and GPS data to develop a graph based SLAM in order to generate millimeter precise geo-referenced offline point cloud based map for the targeted devices to achieve accurate Navigation.
• Designed and implemented a point cloud and GPS based Localizer for precise real-time localization.
• Executed the implementation of a range of grid-based and sampling-based path planning algorithms.
• Developed and deployed path tracking algorithms tailored for low-speed autonomous robots.
• Implemented real-time obstacle detection and multi-object tracking systems to ensure safety and enhance situational awareness in autonomous environments.
• Engineered a mobility platform for robots and forklifts, enabling joystick-based control. Implemented CAN message translation for efficient device operation and maneuverability.
• Created a user-friendly GUI enabling customers to remotely operate and initiate autonomous driving functionalities with ease.
• Contributed to the setup and integration of sensors and collecting data for cars, drones, forklift, and robots.

C++, Python, PCL, OpenCV, Eigen, ROS, g2o, Ceres, Git, Docker, Raspberry Pi, Jetson Nano, Jetson Xavier

Computer Vision Engineer

April, 2020 - November, 2020

SONAH GmbH, Aachen Germany

• Developed depth quantized custom lightweight object detection model for inter-frame tracking on embedded devices.
• Engineered an inter-frame multi-object tracker to enhance object tracking accuracy.
• Designed a depth quantized lightweight segmentation model for efficient segmentation tasks.
• Developed a statistical data analysis tool for object detection and segmentation to mitigate model biases.
• Engineered a custom object detection model tailored for perspective transformation-based region-of-interest targeting embedded devices.
• Enhanced existing classifiers through the implementation of case-oriented augmentation schemes.

C++, Python, OpenCV, Keras, Tensorflow, Tensorflow Lite, Git, Docker, Raspberry Pi

Computer Vision Engineer (Internship)

March, 2020 - March, 2020

SONAH GmbH, Aachen Germany

• Developed custom object detection model for perspective transformation based region-of-interest for embedded device

Python, OpenCV, Tensorflow, Raspberry Pi

Data Scientist (Work Student)

April, 2019 - September, 2019

Teraki GmbH, Berlin, Germany

• Implemented live streaming of compressed videos from embedded devices to web browsers, focusing on region-of-interest selection using Neural Networks.
• Contributed to various region-of-interest detection schemes leveraging neural networks.
• Documented research papers detailing the findings and methodologies.

C++, Python, OpenCV, FFmpeg, Keras, Tensorflow, Flask, NVIDIA Jetson Nano, Raspberry Pi

Data Scientist (Internship)

September, 2018 - March, 2019

Teraki GmbH, Berlin, Germany

• Contributed to the development of various region-of-interest detection schemes utilizing feature descriptors.
• Developed a Windows application using C++ to automate the compression of 4K and 1080p videos, focusing on region-of-interest selection.

C++, Python, OpenCV, FFmpeg, Qt Creator, Visual Studio, Raspberry Pi

Junior Manager

July, 2012 - March, 2014

Infrastructure and Energy Services Ltd, Mumbai, India

• Designed power flow diagrams, wiring diagrams, and electrical schematics for distributed substations.
• Prepared scope of work, cost estimation, and procured materials for distribution substations.
• Coordinated technical aspects with clients and design departments.
• Monitored weekly progress of electrical and instrumentation activities.

Matlab, ETAP, AutoCAD