I am interested in the fusion of technology with the real world, especially in the field of Data Science. I love to get exposed to novel ideas and figure out ways to bring them to life. I aspire to make efforts that produce visible effects within the community.
I tweet @Satvik_Sh
I tweet @Satvik_Sh
SKILLS
- Languages - Python, Java, JavaScript, Rust, R, SQL, C++, MATLAB
- Frameworks - TensorFlow, pyTorch, scikit-learn, Spring, Flask, Falcon, React
- Tools - Spark, Azure, AWS, Docker, Unix/Linux, Latex
EDUCATION
University of Washington, Seattle - Bachelor of Science
Junior - Computer Science, GPA: 3.81
Annual Dean's List
Coursework:
University of Washington, Seattle - Bachelor of Science
Junior - Computer Science, GPA: 3.81
Annual Dean's List
Coursework:
- Deep Learning
- Advanced Machine Learning
- Data Structures and Parallelism
- Cryptography
- Programming in R
- Quantitative Statistics for Data Science
- Statistical Machine Learning
- Data Management
- Intro to Algorithms
- Database Internals
- Natural Language Processing
- Advanced Multivariable Calculus
- Linear Algebra
- Differential Equations
- Vector and Complex Analysis
- Scientific Computing and Numerical Analysis
- Web Development
- Art History: Renaissance to Modern Art
- Art History: Michelangelo
- History of Immigrants in the U.S.
- Art, Architecture, and Society in London
- Contemporary Britain
- London's Contemporary Theater
PROJECTS
Data Constrained classification using Residual Networks (2018)
Data Constrained classification using Residual Networks (2018)
- Developed a novel training approach to deal with classification problems in data constrained environments (~100 samples per class).
- Achieved 4x improvement in validation accuracy over vanilla transfer learning by using greedy training, auxiliary loss, and elastic loss regularizer.
- Stack: Python, keras, pandas
- Developed evaluations and wrote a paper to perform imbalanced learning on Airbnb's Dataset and predict the country of booking for a new user.
- Stack: Python, XG Boost, scikit-learn (AdaBoost, Losgistic Regression, MLP, and more), numpy, pandas
- Developed a tool to curtail the rise of stolen packages from the porch of residential buildings/apartments using security camera feed.
- Used Object Detection to generate tags for each frame of input through Cognitive Services and inform user of suspicious activities using Twilio.
- Stack: Python, OpenCV, scikit-image, Microsoft Cognitive Services (Computer Vision), Twilio API
- Developed a platform to contrast top GitHub repositories based on 'Issues comments' with any user-input GitHub repository.
- Used Watson's Tone Analysis over the comments extracted from most the active repositories to quantify standards for open-source collaboration.
- Stack: Python, IBM Watson, Github API, bokeh
- Analyzed the emergence of patterns across different color spaces in over 150 paintings from the 15th century using Image Analytics.
- Stack: R, Shiny, highcharts, Tableau
- Surveyed the education indicators among different countries from the MDG dataset published by the UN for past 20 years.
- Using Bing Maps API, R, JS, CSS, and HTML, made a web application that used Seattle Police Department's records to re-route users to take the safer routes instead of the quicker ones.
- Stack: R, Bing Maps API, HTML, CSS, JavaScript
- Developed a Microsoft HoloLens application to render solutions to input virtual equations and questions.
- Stack: C#, Wolfram Alpha API, Unity
- Developed a chat bot to help and provide solutions for countering the growing rate of depression among teenagers.
- Digitized the process of filling out of forms during medical emergencies by making wireframe and prototype of a web application, thus saving valuable time.
EXPERIENCE
Deep Learning (CSE 599G1), Allen School @ University of Washington - Teaching Assistant (September 2019 - Present)
Deep Learning (CSE 599G1), Allen School @ University of Washington - Teaching Assistant (September 2019 - Present)
- Hold office hours every week to help students with assignments/concepts as discussed in the class.
- Develop docker containers to grade weekly assignments of over 150 students in isolated environments.
- Code and test machine learning algorithms to be added to the package for reference and use in other projects.
- Developed, deployed, and tested a tool to reduce the problem resolution time for Hulu Live TV customers in cases of wrong channel listing.
- Designed a backend API to consume multiple microservices and decode relationships between geographic and channel metadata.
- Integrated the backend with an interactive frontend to facilitate easier conflict resolution for Live TV users by Resolution Specialists.
- Technologies used: Java (Spring, RxJava, Lombok, Mockito), React, Leaflet
- Led the project development and research on a Machine Learning based interactive art exhibition for the Harvard Art Museums.
- Developed and trained deep ensemble models and GANs to learn about the evolution in temporal settings.
- Designed and implemented scalable API with containerized database, backend, and frontend to render predictions from models.
- Headed a workshop to give an introduction to Data Visualization, Machine Learning, Deep Learning, and GANs.
- Technologies used: Python (pyTorch, scikit-learn, XGBoost, pandas, falcon, flask), PostgreSQL, Docker
- Explored the existing techniques in Few Shot Learning and trained models to make progress towards better generalization.
- Conducted experiments surrounding Snapshot Ensembles, Autoencoders, and Superpixel segmentation to discover their trade-offs.
- Led a discussion section for 30 students every week, held office hours, and graded weekly assignments/examinations.
- Topics involved: Probability, Random Variables, Central Limit Theorem, and Maximum Likelihood Estimation.
- Organizeed workshops twice every week to teach students course content from higher level linear algebra and differential equations.
- Guided and tutored students on concepts of calculus and pre-calculus to enhance the problem solving skills and refine the fundamental concepts.
- Assisted head faculty member with classroom instruction material and monitor student progress to identify areas that require improvement.
- Technologies involved: Omeka, Digital Scholar Lab, OpenRefine, Lexos, Voyant
- Developed a Neural Network based off ResNet-50 to infer identity of artists from photographs achieving over 40% accuracy.
- Implemented Neural Style Transfer to make the user-input images resemble a painting from the predicted lesser-known artist.
- Libraries and resources used: keras (w/ tensorflow backend), Amazon Elastic Beanstalk, flask, wikipedia, numpy, pandas, imagesoup, JavaScript, kaggle (dataset)
- Presentation
- Presentation Slides
- Code
- Pioneered the transformation of the existing websites to make them easily accessible to research teams and potential users. Technology: Drupal
- Transcribed, summarized, and revised texts from ancient diaries that offer a glimpse into the life, trade, and politics of the Ottoman Baghdad.
- Developed and administered plans for effective organization of a hackathon with over 150 participants.
- Managed and coordinate the creation and management of the web-content for effective communication and exposure.
- Volunteered twice a week as Barista and Cashier to help-out with day-to-day operations.
- Examined the behavior of the website audience and identified ways to increase the interaction time and improve the user experience through SEO. Technology: Google Analytics
- Communicated with customers, employees, and other individuals to answer questions, disseminate or explain information, and address complaints.
- Managed the facility of the day-care-center for homeless women and provided IT support.
- Coordinated events and ensured the smooth operation of the $120,00 grossing fundraiser.
AWARDS
- Best in Show - SWEHacks (2018)
- Phi Sigma Theta National Honor Society Nomination (2018)
- Best in Show - DefHacks (2017)
- Best Potential for Social Impact - Foundation of Informatics (2017)
- Golden Key International Honors Society Nomination (2017)
- Society of Collegiate Leadership and Achievement Nomination (2017)
- AP Scholar with Distinction (2016)
- AP Scholar (2015)
RECOMMENDATIONS
Podcasts
Podcasts
- The Daily
- Today, Explained
- So Money
- Bombshell
- Analysis
- Listen Money Matters
- Left, Right & Center
- The Minimalists
