Savinay Shukla

My journey into the world of technology began at a young age when, at just 8 years old, I experienced the thrill of independently installing my first operating system. This early fascination ignited a deep passion within me for computers and their endless possibilities. The sense of accomplishment and curiosity that accompanied that first successful installation laid the groundwork for a lifelong dedication to the field.

From disassembling computers to constructing complex deep learning applications, my expertise and knowledge grew exponentially. Time seemed to fly by, and now, in the year 2023, I find myself in the bustling city of New York, driven to expand my skill set and knowledge even further. Currently, I am pursuing a Master's degree at New York University in Computer Engineering, immersing myself in advanced technologies and embracing the challenges and opportunities that lie ahead.

1. Aug '23 — Present

   Graduate Teaching Assistant · NYU Courant Institute of Mathematical Sciences

   Graduate Teaching Assistant

   New York University, New York City

   Spearheading support in implementing advanced deep learning systems within distributed environments using High-Performance Computers (HPCs).

   • Python
   • Pytorch
   • High-Performance Computing
   • Tensorflow
   • LLMs
   • Computer Vision
2. May '23 — Aug '23

   Section Leader · NYU Center for Data Science

   Graduate Adjuct Employee

   New York University, New York City

   As a lab section leader for the DS-UA 301: Advanced Data Science course under the guidance of Professor Parijat Dube, I assumed the crucial responsibility of conducting practical lab sessions. I guide students in the implementation of Machine Learning frameworks like Scikit-Learn, Tensorflow, and Pytorch, ensuring their understanding and practical application of these tools.

   • Python
   • Tensorflow
   • Scikit-Learn
   • Pytorch
3. Jan '23 — May '23

   Course Assistant · New York University

   Graduate Course Assistant

   New York University, Brooklyn.

   I assisted Silver Professor David J. Pine in curating course content for his course of Computation Chemistry in the Chemical and Biomolecular Engineering School of NYU Tandon.

   • Computational Chemistry
   • Python
   • Numpy
   • Numba
4. Dec '19 — Aug '22

   Software Engineer · IBM

   IBM India

   Bangalore, India

   • Responsible for the development and deployment of API improvements primarily focused on application migration.
   
   • Leveraged Java 8, Hibernate, and Oracle Database to create a robust and efficient backend for the microservices
   
   • Boosted database read performance by analyzing queries, optimizing key columns for indexing, and achieving a 30% reduction in read time for improved efficiency
   
   • Collaborated within an agile framework to perform code reviews and quality assurance, improving overall code quality by 20%.
   
   • Awarded “IBM Gold Champion Learner - 2020” recognition for a continuous learning initiative.

   • Certifications & Badges

   • Java
   • Spring Boot
   • JavaScript
   • TypeScript
   • Oracle
   • Node.js
   • SQL
5. Feb '19 — Jun '19

   Data Science Intern · Morning Blaze Pvt. Ltd.

   Start-up

   Pune, India

   • Engineered a data scraping pipeline for extracting market indicators (commodities, forex, global markets), enhancing predictive modeling for BSE automotive stock opening prices.
   
   • Implemented lightweight time forecasting models, integrating market and technical indicators to boost accuracy.
   
   • Conducted comprehensive research on historical data, ensuring incorporation of extensive indicators to effectively capture stock market downturns
   
   • Optimized the engine across 300+ BSE companies, resulting in an impressive 5% reduction in prediction losses

   • Python
   • Finance
   • LSTMs
   • Global Indices

• World on Wheels

  Engineered a responsive car-rental front-end using Angular and TypeScript for an intuitive interface.
  Integrated TailwindCSS for streamlined, visually appealing design, enhancing overall user experience.
  
  Prioritized user-centric design, incorporating authentication and historical booking records for a personalized experience.

  • Angular
  • TailwindCSS
  • TypeScript
  • MySQL
  • Django

  
• Distributed Dual Discriminator GANs

  Proposed to add another discriminator in a typical DCGAN training pipeline for the Generator to acheive better and faster convergence. Achieved a 40% speedup in time to achieve optimal FID and IS Scores across various datasets which include CIFAR-10, CIFAR-100, MNIST and SVHN.
  
  Additionally, extended this implementation on a distributed setting using Pytorch's DDP to know possible limitations and bottlenecks.

  • Generative Models
  • Pytorch
  • Distributed Data Parallel

  
• Lighweight ResNets with Squeeze and Excitation

  A Lightweight ResNet-18 implementation that has less than 5 million trainable parameters as opposed to the original implementation with 11 million parameters. Leverages channel-wise attention mechanism called Squeeze and Excitement
  
  Model achieves an accuracy of 92.17% within 70 epochs on CIFAR-10.

  • Pytorch
  • ResNets
  • Attention Mechanisms

  
• ClearView - Depth-wise Seperable CNNs

  Our approach employs depth-wise separable convolutional layers and self-regularized activation functions to effectively clean hazy images and can be integrated with segmentation models for enhanced performance.
  
  We believe that our approach will be particularly useful for memory- bound systems already confined by heavy segmentation models when dealing with obscured environments.

  • Pytorch
  • CNNs
  • Depth-wise Seperable
  • DehazeNets

  
• Just Breathe - Detecting Breathing Anomalies.

  Leveraging the power of ARM Cortex-M4 and BNO055 Orientation Sensor, designed a real-time embedded system on a STM32 Discovery board to detect breathing anomalies of patients.

  • C++
  • PlatformIO
  • ARM Cortex-M4
  • STM32

  
• C-Match - Real-Time Score Tracking

  A Microservice-based web application built using Angular, Spring Boot, MySQL, MongoDB, and Docker. Enables users to stay updated with live cricket scores and statistics in real-time.
  
  The application provides a responsive and interactive user interface using Angular Material. Dockerization ensures easy deployment and scalability.

  • Angular Material
  • Spring Boot
  • MongoDB
  • MySQL

  
• Market Watcher - Recommendation Engine

  Developed a stock market recommendation engine using Pandas and Keras. Incorporated global market indices and technical indicators as prominent features to reduce prediction errors. Optimized engine across 300+ BSE companies, resulting in reduced prediction losses of up to 2%.

  • Python
  • GANs
  • Keras
  • Finance