devrobotics
I am a Robotics Systems Engineer at Amazon Robotics, where I work on deploying large-scale robotic systems that power Amazon’s fulfillment and logistics network.
I'm also building Polaris, a personal assistant that runs on a Raspberry Pi using OpenClaw, helping me edit this site, track changes, and automate my workflows.
Beyond my current role, my path into robotics looks like this:
During my master’s at Northeastern University, I worked on an autonomous UAV research project in collaboration with Prof. Rifat Sipahi:
A Robust Object Classification System that processes video frames to classify 15 different objects from a live webcam video stream. Leverages K-Nearest Neighbor Classification to achieve 95% accuracy in object classification.
C++, OpenCV, XCode
An industrial grade autonomous UAV spanning 1000mm capable of producing over 22lbs of thrust equipped with several sensors, capable of surveying, static obstacle avoidance, target classification and air delivery.
C++, Python, OpenCV, PX4, MavLink, MavProxy, MavROS, Pixhawk 2.4
Conducted a comparative analysis of optical flow estimation techniques, evaluating the accuracy of classical approaches such as the FarneBack algorithm, and deep learning-based methods like FlowNet 2.0. Evaluated endpoint error, angular error, flow discontinuity, velocity estimation, object tracking, pixel displacement and performed bounding box analysis.
C++, Python, OpenCV, XCode, OpenGL, PyTorch, TensorFlow
These are some of my mini projects, that helped me master some concepts and helped me build practical skills in various areas of technology and development. Each project represents a unique learning experience and showcases my ability to apply theoretical knowledge to real-world scenarios.
An easy way to view all my projects is my LinkTree
The objective of this project is to generate a panoramic or mosaic image by stitching multiple images together. This is achieved through a series of steps, including camera calibration to remove distortions, detecting major features using the Harris Detector, and stitching images with overlapping Harris corners. The end goal is to create a seamless and distortion-free panorama from the input images.
Skills Honed:Camera calibration, distortion correction, feature detection, Harris Detector, image stitching, image transformation, panorama creation.
This project aims to apply real-time image filtering to webcam input. It involves several image processing tasks, such as converting images to grayscale, applying Gaussian blur, cartoonizing with gradient magnitude and blur/quantize filters, and adding salt and pepper noise. The core concept revolves around using convolution techniques to manipulate images. The desired image manipulations are achieved through pixel-level operations.
Skills Honed:Real-time image processing, webcam input, grayscale conversion, Gaussian blur, cartoonization, gradient magnitude, convolution, noise addition.
This project aims to retrieve images from a dataset based on their characteristics, such as color, texture, and spatial layout, using a selected target image. It involves the extraction of feature vectors and the calculation of the top N matches, providing practical experience in image matching and pattern recognition.
Skills Honed:Content-based image retrieval, feature vector extraction, color spaces, histograms, texture features, spatial layout, distance metrics, pattern recognition, Gabor filters, parameter tuning.
This project involves camera calibration and processing image sequences from a webcam to detect and isolate a chessboard pattern. It projects detected points as 3D points for tracking movement and uses robust features. This application extracts and displays chessboard corners, performs camera calibration, estimates camera pose, and overlays a virtual object on the image frame. It also extends to perform perspective transformation and blending of images.
Skills Honed:Camera calibration, feature detection (SURF Features), image processing, pose estimation, virtual object overlay, perspective transformation, QR code detection, Augmented Reality techniques.
The objective of this project is to set up and collect RTK GPS (u-blox ZED-F9P) data from both a base station and a rover. This data collection is conducted in open spaces and locations with occlusions to analyze and compare the gathered data. The goal is to eliminate common noise, study the differences between stationary and moving data, and draw meaningful inferences from the collected data sets.
Skills Honed:Hardware setup for RTK GPS data collection, base station and rover configuration, noise elimination, stationary and moving data collection, data comparison, analysis, and drawing inferences from the collected data.
The project's objective was to configure the VectorNav IMU hardware to collect data in the $VNYMR format and set the data collection rate at 40Hz. Data collection took place in stationary conditions, and various parameters including Orientation, Angular Velocity, Linear Acceleration, Magnetic Field, Noise Characteristics, Error Distribution, and Allan Deviation were analyzed for the IMU data.
Skills Honed:Configuration of IMU hardware, data format setting, data collection rate adjustment, data analysis of parameters including Orientation, Angular Velocity, Linear Acceleration, Magnetic Field, Noise Characteristics, Error Distribution, and Allan Deviation.
Feel free to reach out if you'd like to discuss exciting opportunities in the world of robotics, computer vision, and drones, or if you're interested in collaborating on groundbreaking projects with a passionate robotics enthusiast like me.
