Cambridge, Massachusetts, United States
Passionate Data Scientist with a keen interest in turning complex data into actionable insights. Skilled in Statistical Analysis, Deep Learning, Computer Vision, LLMs and Data Visualisation. Eager to contribute my analytical mindset and coding proficiency to tackle real-world challenges and drive informed decision-making. Always open to the new learnings and opportunities to grow in the dynamic field of Data Science.
I develop and lead scalable automation and machine learning frameworks that enhance data workflows, accelerate scientific reporting, and support data-driven decisions across cross-functional teams.
• Reviewed 30+ research papers for a Proof-Of-Concept project, and conducted statistical analysis on temporal energy data at hourly interval. • Developed an ETL pipeline using Amazon S3 bucket and AWS Lambda function to develop a Data Warehouse for real-time visualization. • Built Power BI dashboards, carried out multivariate time-series forecasting, and simulated energy models to analyse building performance.
CV Lab | Advisor: Dr. Haibin Ling Conducted research on Self Supervised Machine Learning models to develop state-of-the-art techniques for Image denoising and particle picking (image segmentation) of protein cells from 2D images obtained through Cryo Electron Microscope.
Robotic Innovations Lab | R&D | Deep Learning | Computer Vision | 3D Modelling • Built 3D Model generation pipeline by stitching multiple point clouds of an object using registration algorithms. Also, used multiple softwares for object model post-processing such as smoothing mesh using Laplacian Smoothing filter, Cleaning and Repairing Mesh using Surface Reconstruction Poisson filters, computing Vertex Normals, etc. to generate perfect 3D models. • Created an end to end pipeline for synthetic dataset generation using 3D object model for deep learning training tasks. The Blender physics engine is being used to place objects on the surface by randomly dropping them in the space. • Built 6DoF object pose estimation algorithm by training deep-learning based FFB6D algorithm for various objects. The predicted pose is then fed to the object grasping algorithm to assist the robot in picking the objects. • Built object segmentation algorithm by training Mask RCNN algorithm on various datasets. Integrated this algorithm with 6D pose estimation pipeline to create an end to end application for object detection and object pick & place. • Worked with various sensors such as Realsense D415, Realsense SR300, IDS Ensenso camera and Kinect PrimeSense for 3D scene capturing. • Trained various deep learning models with single object dataset of up to 80000 examples, tuned model hyper-parameters and analysed trained models. Also, applied preventive measures for model overfitting & underfitting. Skills: Computer Vision · Deep Learning · Data Science · Data Visualization
Analysed time series data and utilized RNNs to forecast the opening stock price for the following day, on the basis of the closing price of the stock for the current day. Trained on Stock data for past 30 years with 10,000 examples. Accuracy of around 80% was achieved.