Hrushikesh Chavan

Experience

• 

  Argus System

  Jul 2016 - Dec 2017

  Data Science Engineer
• 

  Indian Institute of Technology, Kanpur

  Jan 2018 - Oct 2020

  Machine Learning Researcher

  Developed Open Source Python-Based Tool for Generating STL Files of TPMS Surfaces for 3D Printing Developed a Python-based tool that enhanced the generation of complex 3D printed structures by applying advanced algorithmic techniques and leveraging open-source libraries, significantly improving the efficiency and practicality of 3D printing Objective: Designed and implemented an open-source tool to generate STL files of Triply Periodic Minimal Surfaces (TPMS) suitable for 3D printing, enhancing the ability to create complex and efficient structures.Algorithm Used: Leveraged the Lewiner Marching Cubes algorithm for efficient and accurate isosurface extraction from volumetric data.Technical Implementation:Implemented the gyroid function to define the mathematical representation of TPMS.Utilized numpy for creating a 3D grid and skimage.measure for applying the Marching Cubes algorithm.Generated vertices and faces for the isosurface and created a mesh structure using the numpy-stl library.Enabled the addition of thickness to the generated surfaces by creating offset surfaces, ensuring robust and practical 3D printed structures.Outcome: Successfully generated STL files of TPMS structures with adjustable thickness, facilitating advanced 3D printing applications and contributing to the open-source community.Technologies Used: Python, Numpy, Scikit-Image, Matplotlib, Numpy-STL. Show less

• 

  The University of Tokyo

  May 2019 - Jul 2019

  Data Science Researcher

  Supervisor: Dr Hiromasa Suzuki, Professor, Precision Engineering Department, University of TokyoTitle: 3D printing of Different shapes with infill of Schwarz D Surface• Developed program to generate stereolithography (STL) file of Triply Periodic Minimal Surfaces • 3D printed the Schwarz D Surface• Developed program to generate STL files of different shapes with infill of TPMS surface and 3D printed it by Z250 printer.• Software used: Python, MATLAB, OpenSCAD, MeshLab Show less

• 

  DMG MORI

  Jun 2019 - Jul 2019

  Research And Development Engineer

  Department: Next Generation Machine Development Department, LT6600 3D GroupSupervisor: Mrs Yoko Hirono, Assistant Chief Engineer, R&D HQ, Next Generation Machine Development Department, LT6600 3D GroupProject Title: Design and Analysis of Jig for Transportation of Cable Control Unit of LASERTEC 6600 3DObjectives of Transportation Jig of Cable Control Unit:1) The jig should support Cable Control Unit during Transportation.2) It should resist static forces exerted by Cable Control Unit.3) It should withstand forces due to acceleration during transportation.4) It should resist Seismic forces occurs during the earthquake.I did analysis to check static deflection so that results obtained can compare with Theoretical deflection.I did analysis considering forces due to acceleration during transportation and also verified resistivity of jig against seismic forces.Material of Jig: STKR 400 Purpose: This material easily available in square and rectangular bars.Software used for Modelling and Analysis: Creo Parametric 4.0 Show less

• 

  Eaton

  Oct 2020 - Mar 2023

  AI/ML Engineer

  1. Enhancing User Experience through NLP:Accomplished: Developed an NLP model to discern user sentiments, measured by a 20% increase in user engagement.Methodology (Y): Analyzing user reviews with advanced NLP techniques.Impactful Action (Z): Executed comprehensive text preprocessing and exploratory data analysis, employing techniques such as lowercasing, stop word removal, stemming, lemmatization, and tokenizer. Implemented text vectorization methods like Bag of Words and Tf-IDF. Utilized Word Clouds for insightful exploratory data analysis.Result: Identified key areas for product and service improvement, contributing to enhanced user satisfaction.2. Aerospace Application Simulation Efficiency:Accomplished: Developed a transfer function to predict hose fitting failure, resulting in a 60% reduction in data acquisition time.Methodology (Y): Utilizing design of experiment (DOE) methods for efficient simulation.Impactful Action (Z): Generated data points for simulation, saving 60% of the time previously required for data acquisition. This led to a significant reduction in the time needed for new product launches.Result: Improved efficiency in aerospace application simulations, ensuring timely product launches and reduced costs.3. Boosting Conversion Rate in Aerospace Sales:Accomplished: Improved conversion rate from 30% to 70% for new enquiries in the aerospace sector.Methodology (Y): Developing a model to identify promising enquiries with high conversion potential.Impactful Action (Z): Managed null values in categorical and numerical columns using techniques like Simple Imputer and KNN Imputer. Implemented dimensionality reduction techniques such as PCA, Variance Inflation Factor, and p-value analysis.Result: Successfully increased the conversion rate, optimizing the sales process and enhancing revenue generation. Show less

• 

  John Deere India Pvt. Ltd. (JDTCI)

  Apr 2023 - Oct 2024

  Senior Data Scientist

  Gen AI-Powered Visual Chatbot for Rapid AssistanceDeveloped an advanced system to generate videos from documents. Utilizing Multimodal RAG, the system dynamically produces video responses by integrating contextually relevant images and synchronized audio. This solution leverages sophisticated image retrieval and text-tospeechsynthesis, enhancing information accessibility and engagement, while optimizing content creation workflows and reducing operational costs for businesses. Multimodal RAG, LlamaIndex, Lance DB, GPT 4oSurvival Analysis for Product Reliability:Achieved an 83% reduction in runtime for product reliability analysis using Survival Analysis (Weibull Distribution) by developing and deploying efficient model-building code.Measured success through enhanced workflow and productivity, setting a new benchmark for automated life data analysis.Successfully eliminated software dependencies, resulting in a substantial cost savings of USD 65,000 in license fees.Software Fault Prediction and Risk Mitigation:Pioneered a comprehensive risk quantification and mitigation approach throughout various phases of the software development lifecycle.Implemented a machine learning-based software defect prediction model, reducing development costs by detecting faults earlier.Improved software quality, reliability, and efficiency through the creation of a robust bug prediction model using ML algorithms.Demonstrated expertise in understanding key factors influencing software performance.Document Focused Chatbot:Innovated and developed a chatbot tailored for document-related queries, significantly enhancing user experience.Utilized advanced techniques such as embeddings using sentence structure to enable the chatbot to extract relevant information from documents.Transformed the user interaction experience by efficiently retrieving important information from documents and providing precise answers. Show less

• 

  Revvity

  Nov 2024 - now

  Principal AI Engineer and Data Scientist