1<!DOCTYPE html> 2 3<!-- 4 portfolYOU Jekyll theme by yousinix 5 Free for personal and commercial use under the MIT license 6 https://github.com/yousinix/portfolYOU 7--> 8 9<html lang="en" class="h-100"> 10 11<head> 12 13 14 15 16 17 18 19 <meta charset="utf-8"> 20 <meta http-equiv="X-UA-Compatible" content="IE=edge"> 21 <meta name="viewport" content="width=device-width, initial-scale=1"> 22 23 <meta property="og:type" content="website"> 24 <meta property="og:title" content="Work"> 25 <meta property="og:description" content="I turn tea :tea: into code, Want to discuss ?? Always up for Robotics, Artificial Intelligence, Quantum Physics, Neuroscience & Life!!! ..."> 26 <meta property="og:image" content="./assets/uday.png"> 27 28 <title>Work</title> 29 <meta name="description" content="I turn tea :tea: into code, Want to discuss ?? Always up for Robotics, Artificial Intelligence, Quantum Physics, Neuroscience & Life!!! ..."> 30 31 <link rel="shortcut icon" type="image/x-icon" href="/assets/favicon.ico"> 32 33 <!-- Theme style --> 34
34<script src="/assets/js/theme.js"></script>
34 35 36 <!-- Font Awesome CDN --> 37 <link rel="stylesheet" href="https://use.fontawesome.com/releases/v5.15.4/css/all.css"> 38 39 <!-- Bootstrap CSS CDN --> 40 <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/css/bootstrap.min.css"> 41 42 <!-- Animate CSS CDN --> 43 <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/animate.css/3.7.0/animate.css"> 44 45 <!-- Custom CSS --> 46 <link rel="stylesheet" href="/assets/css/style.css"> 47 48</head> 49 50 51<body class="h-100 d-flex flex-column"> 52 53 <main class="flex-shrink-0 container mt-5"> 54 <nav class="navbar navbar-expand-lg navbar-themed"> 55 56 <a class="navbar-brand" href="/"><h5><b>Uday Girish Maradana</b></h5></a> 57 58 <button class="navbar-toggler" type="button" data-toggle="collapse" data-target="#navbarNavAltMarkup" aria-controls="navbarNavAltMarkup" aria-expanded="false" aria-label="Toggle navigation"> 59 <i class="fas fa-1x fa-bars text-themed"></i> 60 </button> 61 62 <div class="collapse navbar-collapse" id="navbarNavAltMarkup"> 63 <div class="navbar-nav ml-auto"> 64<a class="nav-item nav-link " href="/projects/">Projects</a> 65 66 <a class="nav-item nav-link active" href="/work/">Work</a> 67 68 <a class="nav-item nav-link " href="/publications/">Publications</a> 69 70 <a class="nav-item nav-link " href="/courses/">Courses</a> 71 72 <a class="nav-item nav-link " href="/blog/">Blog</a> 73 74 <a class="nav-item nav-link " href="/skills/">Skills</a> 75 76 <a class="nav-item nav-link " href="/about/">About</a> 77 78 <a class="nav-item nav-link " href="/contactme/">ContactMe</a> 79 80 <a class="nav-item nav-link " href="/outreach/">Outreach</a> 81 82 <a class="nav-item nav-link " href="/personal_life/">Personal</a> 83 84 85 86 <span id="theme-toggler" class="nav-item nav-link" role="button" onclick="toggleTheme()"></span> 87 </div> 88 </div> 89 90</nav> 91 <div class="col-lg-10 mx-auto mt-5 markdown-body"> 92 <h2 id="projects---professional-internships"><strong>Projects - Professional /Internships</strong></h2> 93<hr> 94<hr> 95 96<h3 id="at-tiger-analytics"><strong>At Tiger Analytics</strong></h3> 97 98<p><strong>MLOps Framework Dev for Unified DataScience Platform</strong></p> 99<ul> 100<li> 101Led a team of three ML Engineers in developing a Scalable MLOps Framework in Amazon Web Services utilizing Sagemaker capabilities. 102</li> 103<li> 104Developed and deployed a MLOps framework for deploying Datascience models with ease involving services such as Sagemaker, DynamoDB, Event Bridge, SNS. 105</li> 106<li> 107Developed a Organizational flow for the client to onboard new datascientists to get familiar with framework and reduce their model development and deployment time by more than 75 % (Almost from 4 Weeks to 7 Working Days) 108</li> 109</ul> 110 111<p><strong>MLOps Pipeline Dev for Price Elasticity</strong></p> 112<ul> 113<li> 114Led the GCP ML Accelerator Program and helped organization to become a GCP ML partner. My contribution involve explaining the projects and the organization capbility in terms of both Technical capabilities and the successfull projects and business impact. 115</li> 116 117<li> 118Led a team of four ML Engineers in developing a Scalable and experiment reproducible MLOps pipeline for Price Elasticity of SKU's classification. 119</li> 120<li> 121Able to reduce the total models training time by more than 50 percent using parallelised approaches with Pyspark. 122</li> 123<li> 124End-to-End pipeline with orchestrator as Google Composer, Custom container deployment for Feature preparation, Dataproc for running Pyspark jobs to train more than 30000 SKU models in less than 3 hours. 125</li> 126</ul> 127 128<p><strong>IOT - CV -ML Exploration - Internal POC</strong></p> 129<ul> 130<li> Leading the capability development of the firm in the combined space of IOT and CV. </li> 131<li> Developing scalable architecture on AWS with IOT Greengrass and Kinesis. </li> 132<li> Currently in POC phase where we are trying to work with edge devices with two streams of 133data. One of the stream is a continuous value data (IOT sensors) and one other stream is 134a video stream. </li> 135<li> Exploring the Scalable CV pipelines both in DS and MLOps perspective. Checking the 136feasibility of optimisation of the pipelines from the Data preprocessing of streams to 137Inference. (Elastic Inference, Nvidia Deepstream SDK, Triton Inference Server) </li> 138</ul> 139 140<p><strong>Login Event</strong> <strong>Anomaly Detection - ML in Cybersecurity</strong></p> 141<ul> 142<li>Developed a Feature Engineering and Ingestion Pipeline using MLOps approach 143</li> 144<li>Developed a Feature Serving and Training pipeline using GCP Vertex AI MLOps 145</li> 146<li> eveloped an Online Serving and Eval pipeline using GCP Vertex AI. 147</li> 148<li> Automating the entire ML Dev and Deployment field using CI-CD with GCP services. 149</li> 150<li> Developed models related to Anomaly detection using a hybrid approach of Unsupervised and Supervised models with an additional component of Explainable AI. </li> 151</ul> 152 153<hr> 154 155<h3 id="as-a-freelancer"> 156<strong>As a Freelancer</strong>:</h3> 157 158<p><strong>Project based Freelancing - Client through LinkedIn</strong></p> 159 160<ol> 161 <li>Point Tracking for Ultrasound Videos to measure joint instability 162 <ul> 163<li> Developed a end to end framework for aiding Medical practitioners to track different segments in ultrasound at a point level throughout the timeframe of a video</li> 164<li> This tool can provide real time plots of distance and angle between segments in 2D space and also gives the capability of doing further analysis with a single time run data. </li> 165</ul> 166 </li> 167 <li>Knee Arthroscopy Tool 168 <ul> 169 <li> Developing a workflow with Tradtional and DL based CV to handle Image based relative measurements and transfer them to real world coordinates </li> 170 <li> This tool helps in Surgeries as an aid to Surgeons to get relative real world estimates of the space on which they are performing the surgery <
170/li> 171 </ul> 172 </li> 173 <li>Label Studio Setup for Automated ML Labelling 174 <ul> 175 <li> Developed Code to support Yolov5 as a ML Backend for Automated ML labelling in Label Studio </li> 176 <li> Integrated Yolov5 , OpenMMLab Object Detection models to LabelStudio ML Backend </li> 177 <li> Wrote Scripts to run LabelStudio in a Colab leveraging Colabâs GPU working through a tunneled external url using Ngrok. </li> 178 </ul> 179 </li> 180 <li>Website for Video Analytics 181 <ul> 182 <li>Developing an Website where user can upload videos and select or upload pretrained models (Yolo, CLIP based formats) and get the results streamed on the same website with outputs labelled on the Video. </li> 183 <li> Used Streamlit for Basic frontend, Model Runner API on backend which runs scripts according to the selection on UI. </li> 184 <li>Plans are there to upgrade this to a React based Website with more model or Data format supports. </li> 185 </ul> 186 </li> 187</ol> 188 189<p><strong>Full-time Machine Learning Engineer- Deep Judge:</strong></p> 190 191<ol> 192 <li>Document Text Extraction and Summarisation 193 <ul> 194 <li> Developed an API that can extract text from a PDF and group the pages based on Page start and ending based on OpenCV-based logic, Block detection, and by tesseract. </li> 195 <li> Used Hugging Face - T5 Transformer inference for Summarisation of the extracted text. </li> 196 </ul> 197 </li> 198 <li>Exploration of NER (Named Entity Recognition) Models (SOTA) and datasets.</li> 199 <li>Information extraction from .docx (XML) formatted documents through traditional python libraries for further NER and NLP-based tagging pipeline.</li> 200 <li>Exploration on SPARQL and Semantic Web based approaches for graph creation for linking the content.</li> 201</ol> 202 203<p><strong>Part-time Data Science Consultant - Aays Analytics:</strong></p> 204 205<ol> 206 <li>Retail Analytics Forecast : 207 <ul> 208 <li>Developing a Proof of Concept Forecast model for Retail Fashion Products (Online mode of selling) to help the planning team about stock intakes, the product should be restocked or not. 209 </li> 210 <li>Experimenting with multiple algorithms and problem statements is solved by modeling it as a Classification + Regression Problem.</li> 211 </ul> 212 </li> 213</ol> 214 215<p><strong>Part- time Project - Applied Computing</strong></p> 216 217<ol> 218 <li>Auto Ticket Generation POC: 219 <ul> 220 <li> Working on developing a platform to automate ticket generation for queries and complaints </li> 221 <li> End to End platform to automate the Phone calls/messages received by a contact center and use AI for automatic ticket generation </li> 222 </ul> 223 </li> 224 <li>Virtual KYC: 225 <ul> 226 <li>Developed a full-fledged Virtual KYC solution with OpenCV, AWS Textract, and Facerec APIs to extract information, verify faces and textual information, and use it for the KYC authentication system. Here the solution provides the total information extracted plus whether the documents are acceptable or not or any discrepancies in the documents.</li> 227 </ul> 228 </li> 229</ol> 230 231<p><strong>Part-time Project- News to WordPress</strong></p> 232 233<ol> 234 <li>Multi languages news translator</li> 235</ol> 236<ul> 237<li> 238 Developed a Fast API + Streamlit based Web interface which can take an e-news or normal paper image picture and process it using Vision API/ Open source-based Pytesseract to extract information (Block level), detect language, and translate it to a set of languages according to the client specification and publish it to different WordPress domains as a new Post. </li> 239 </ul> 240 241<p><strong>Client through Connections:</strong></p> 242 243<ol> 244 <li>Reinforcement learning-based trading system: 245 <ul> 246 <li>Worked on developing codebase to develop a base Reinforcement learning-based trading framework taking the help of opensource work. Especially integration Stable Baselines3 to the OpenAI Gym framework. Worked on A2C, PPO, DQN, DDPG integration to the current system using the CNN and MLP based feature network(Current Default support).</li> 247 </ul> 248 </li> 249</ol> 250 251<p><strong>Upwork (Client Details hidden) :</strong></p> 252 253<ol> 254 <li>Q&A Pipeline: 255 <ul> 256 <li> 257 Developing base question and answering pipeline for demonstration and tutorial purpose using Hugging Face transformers, Pytorch lightning, and model metadata logging to Weights and Biases. </li> 258 </ul> 259 </li> 260</ol> 261 262<hr> 263 264<h3 id="at-new-space-research--technologies"><strong>At New Space Research & Technologies</strong></h3> 265 266<ol> 267 <li> 268 <p>Computer Vision Pipeline for Targeting and Navigation:</p> 269 270 <ul> 271 <li>Worked on developing the current feature matching algorithms (Keypoint extraction algorithms) such as SIFT, SURF which belong to more of the traditional Computer Vision model. </li> 272 <li>So, to increase their accuracy and generalization worked on Deep learning-based feature matching and SLAM-based algorithms such as D2 Net, SuperGlue, and LoFTR which are mostly related to CNN-based or CNN+GNN based approaches.</li> 273 </ul> 274 </li> 275 <li>Platform Shift - Raspberry Pi to Jetson NX: 276 <ul> 277 <li>Lead the exploration of moving the current deployment architecture from Raspberry Pi to CUDA-based platforms such as Jetson Xavier NX and Jetson Xavier AGX to gain a computational advantage in terms of Accuracy and Processing power. </li> 278 <li>Worked on OpenVINO, Triton or TensorRT, Deepstream based deployment methods as a part of this exploration. </li> 279 </ul> 280 </li> 281 <li>Machine Learning- Common Duties: 282 <ul> 283 <li> Working as a part of the ML Engineering team to help convert and optimize the models into Openvino formats and also help in improving the current Object Detection models and participated in training and deployment of models.</li> 284 </ul> 285 </li> 286</ol> 287 288<hr> 289 290<h3 id="at-quantiphi">At Quantiphi</h3> 291 292<ol> 293 <li>Computer Vision for Safety: 294 <ul> 295 <li>This project involves working on computer vision models to build a pipeline to ensure safety in theme park adventure and 4D rides. I was in the Machine learning team of this project and mostly worked on models such as Image Classification, Object Localisation, and Tracking, Pose estimation, and other cloud components to integrate the models into the pipeline.</li> 296 </ul> 297 </li> 298 <li>Computer Vision for Suspect Search: 299 <ul> 300 <li>This project involves working on CV models to build a pipeline where suspects can be searched in the video feed within seconds and to identity thefts and any anomaly activities. I primarily worked in developing Object Localisation and tracking, Person Re-identification, Search logic and clustering logic, and other cloud components.</li> 301 </ul> 302 </li> 303 <li>Document Classification with Transformers: 304 <ul> 305 <li>This project involves working on NLP-based transformer models to integrate the transformer models into a present Kubeflow pipeline and to benchmark different transformer models performance such as Open GPT-2, DistilBERT, Roberta, and Longformer in terms of latency and accuracy.</li> 306 </ul> 307 </li> 308 <li>Translation Service for Document and Web: 309 <ul> 310 <li>This project involves developing a solution that primarily concentrates on translation documents and web pages using Google Cloud APIs (Vision, Translation, Doc Parser) and Auto ML.</li> 311 </ul> 312 </li> 313 <li>Hybrid Model Deployment Scenarios - Solutions research team: 314 <ul> 315 <li>Working on NVIDIA Deep stream SDK, TLT, and connecting different models for optimizing pipeline performance in terms of latency and Accuracy.</li> 316 <li>Working on Tensorflow Lite to understand CPU based deployments and optimizing further to check the feasibility of low-cost model deployments</li> 317 <li>Exploration of other CPU optimizers primarily checking Intel MKL-DNN and Open Vino to understand different CPU level model optimizations.</li> 318 <li>Exploration on Federated learning to understand privacy and distributed model developments.</li> 319 </ul> 320 </li> 321</ol> 322 323<hr> 324<hr> 325<h2 id="roles-responsibilities-and-interests">Roles, Responsibilities and Interests</h2> 326 327<h3 id="professional--level">Professional Level:</h3> 328 329<p><strong>Domain Knowledge:</strong></p> 330 331<ul> 332 <li>Machine Learning</li> 333 <li>Deep Learning</li> 334 <li>Artificial Intelligence</li> 335 <li>Robotics</li> 336 <li>Cloud Computing</li> 337 <li>Edge Computing</li> 338 <li>Quantum Computing</li> 339 <li>Software Engineering</li> 340 <li>Mechatronics</li> 341</ul> 342 343<p><strong>Previous Responsibilities:</strong></p> 344 345<ul> 346 <li>Exploration of SOTA Models</li> 347 <li>Research Paper reading</li> 348 <li>Approaches development for the use case</li> 349 <li>Helping SAâs in Architecture Development</li> 350 <li>Model Development Research</li> 351 <li>Model Development, Training, and Evaluation</li> 352 <li>Model Optimisation (Quantization and Pruning)</li> 353 <li>Model Deployment</li> 354 <li>Integrating model deployments with other cloud components</li> 355 <li>Pipeline regular maintenance metadata scripts</li> 356</ul> 357 358</div> 359 </main> 360 <footer class="mt-auto py-3 text-center"> 361 362 <small class="text-muted mb-2"> 363 <i class="fas fa-code"></i> with <i class="fas fa-heart"></i> 364 by <strong>Uday Girish Maradana</strong> 365 </small> 366 367 <div class="container-fluid justify-content-center"> 368<a class="social mx-1" href="mailto:[email protected]" style="color: #6c757d" onmouseover="this.style.color='#db4437'" onmouseout="this.style.color='#6c757d'"> 369 <i class="fas fa-envelope fa-1x"></i> 370 </a><a class="social mx-1" href="https://www.github.com/udaygirish" style="color: #6c757d" onmouseover="this.style.color='#333333'" onmouseout="this.style.color='#6c757d'"> 371 <i class="fab fa-github fa-1x"></i> 372 </a><a class="social mx-1" href="https://www.instagram.com/udaygirish" style="color: #6c757d" onmouseover="this.style.color='#405de6'" onmouseout="this.style.color='#6c757d'"> 373 <i class="fab fa-instagram fa-1x"></i> 374 </a><a class="social mx-1" href="https://www.kaggle.com/udaygirish" style="color: #6c757d" onmouseover="this.style.color='#20beff'" onmouseout="this.style.color='#6c757d'"> 375 <i class="fab fa-kaggle fa-1x"></i> 376 </a><a class="social mx-1" href="https://www.linkedin.com/in/udaygirish-maradana" style="color: #6c757d" onmouseover="this.style.color='#007bb5'" onmouseout="this.style.color='#6c757d'"> 377 <i class="fab fa-linkedin-in fa-1x"></i> 378 </a><a class="social mx-1" href="https://medium.com/@udaygirish" style="color: #6c757d" onmouseover="this.style.color='#00ab6c'" onmouseout="this.style.color='#6c757d'"> 379 <i class="fab fa-medium fa-1x"></i> 380 </a><a class="social mx-1" href="https://www.twitter.com/UdayGirishM" style="color: #6c757d" onmouseover="this.style.color='#1da1f2'" onmouseout="this.style.color='#6c757d'"> 381 <i class="fab fa-twitter fa-1x"></i> 382 </a><a class="social mx-1" href="https://www.quora.com/profile/Uday-Girish-Maradana" style="color: #6c757d" onmouseover="this.style.color='#ff0000'" onmouseout="this.style.color='#6c757d'"> 383 <i class="fab fa-quora fa-1x"></i> 384 </a> 385 386</div> 387<small id="attribution"> 388 theme <a href="https://github.com/yousinix/portfolYOU">portfolYOU</a> 389 </small> 390 391</footer> 392 393 <!-- Global site tag (gtag.js) - Google Analytics -->
vendor: 66 bytes, lines 393-394
393 394 <script async src="https://www.googletagmanager.com/gtag/js?id=
394UA-175164060-1
vendor: 14 bytes, lines 394-395
394"></script> 395
395<script> 396
vendor: 140 bytes, lines 396-400
396window.dataLayer = window.dataLayer || []; 397 function gtag(){dataLayer.push(arguments);} 398 gtag('js', new Date()); 399 400 gtag('config', '
400UA-175164060-1
vendor: 6 bytes, lines 400-401
400'); 401
401</script>
401 402 <!-- GitHub Buttons -->
403<script async defer src="https://buttons.github.io/buttons.js"></script>
403 404 405<!-- jQuery CDN -->
406<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.3.1/jquery.min.js"></script>
406 407 408<!-- Popper.js CDN -->
409<script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.14.6/umd/popper.min.js"></script>
409 410 411<!-- Bootstrap JS CDN -->
412<script src="https://stackpath.bootstrapcdn.com/bootstrap/4.3.1/js/bootstrap.min.js"></script>
412 413 414<!-- wow.js CDN & Activation -->
415<script src="https://cdnjs.cloudflare.com/ajax/libs/wow/1.1.2/wow.js"></script>
vendor: 1 bytes, line 415
415
416<script> new WOW().init(); </script>
416 417 418<!-- Initialize all tooltips -->
419<script> 420$(function () { 421 $('[data-toggle="tooltip"]').tooltip() 422}) 423</script>
423 424</body> 425 426</html>
Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.