Satvik Bhatnagar Email & Phone Number

Sunnyvale, CA, US

San Francisco Bay Area

• Successfully enhanced high quality image generation, leveraging a variety of advanced diffusion models optimized on top-tier GPUs which resulted in significant improvements in visual fidelity and processing efficiency.
• Focused on Subject Identity Preservation and Virtual Try-On (VITON), conducting extensive research and experimentation with various diffusion techniques. This resulted in high subject image and prompt fidelity.
• Created highly efficient image processing pipelines, combining various trained and pre-trained ML models, reducing the preprocessing time to a few mili-seconds from over 10 minutes.
• Worked on automations that streamlined testing processes and ensured robust deployments.

Irvine, California, United States

• Collaborated with the Chan Zuckerberg Initiative (CZI) to develop a large-scale AI tool using GPT-4 (LLM) API for assessing student responses, analyzing over 150K responses to date.
• Deployed and designed the tool for scalability, enabling daily use by 40K+ students and teachers across 22 schools in the Anaheim Unified High School District.
• Implemented concurrency and manual API requests to optimize a stopping API, reducing analysis time by over 10x.
• Conducting extensive research and testing on LLMs using techniques like Topic Modelling, Fine Tuning, and Prompt Engineering for generating suitable themes from a student's response on an individual as well as class.

Irvine, California, United States

Goa, India

• Developed a Python architecture enhanced with CUDA to identify critical genes that contribute to the growth of certain cancer types, utilizing deep learning techniques on datasets comprising over 100M genetic sequences.
• Designed a sparse deep neural network using Bayesian statistics with spike-and-slab priors, reducing the model complexity by over 80% and overcoming challenges like high dimensions but low sample sizes.
• Achieved feature selection, maintaining an accuracy of over 97%, identifying only the key genetic markers in various synthetic, real-world, and STG Datasets, earning the distinguished A+ Grade as a BTech Project in the.

Japan

• Enhanced the Kaishukahi prediction system by streamlining data processing and employing advanced ML models like Autoencoders and K-means clustering, reducing processing time by 30% and prediction error by over 50%.
• Leveraged MySQL and Apache Hive for efficient data management, ensuring data storage, retrieval, and manipulation.
• Completed unit testing in C++ and Python, achieving 75% code coverage for legacy and new production code.
• Analysed large-scale datasets from over 500M monitors spread across 20 countries and 195 panels, and presented actionable insights such as panel strategies using Power BI dashboards, projecting an increase in annual.

Hyderabad, Telangana, India

• Developed robust NLP models in Python for spoken content information retrieval under non-native contexts, achieving a 25% better retrieval accuracy using techniques like CNNs, RNNs, transformers, auto-encoders, and.
• Implemented techniques like transformers and auto-encoders in the acoustic model and fine-tuning the language model for automatic speech recognition (ASR), resulting in an enhanced transcription accuracy with a 15%.

• Collaborated with the ed-tech startup to visualize KPIs like Revenue Distribution and Course Material Sales, using Tableau, improving revenue tracking by 5x.
• Prioritized and organized the metrics and KPIs for dashboards on the company app and website and provided ideas for new metrics on which user data can be visualized.
• Identified and implemented new metrics for user data visualization, leading to a 150% enhancement in sales visibility, user insights, and engagement with the analytics features for better decision-making.

• Incorporated Neural Ordinary Differential Equations (Neural ODEs) in Weather Prediction, leveraging their efficiency for time series analysis.
• Applied insights from influential papers on Neural ODEs and Maximal Principle Based Algorithms for Deep Learning.

Successfully assisted over 100 students in mastering the Google Cloud Platform through the '30 Days of Google Cloud' Program as a Campus Facilitator at IIT Goa.

Master Of Science - Ms, Data Science, Gpa: 4.0/4.0

Bachelor Of Technology - Btech, Mathematics And Computer Science

Activities and Societies: Co-Founder @ Alpha (Finance Club) | Organizing Team @ CultRang (Cultural Fest), Cepheus (Tech Fest) | Drama Club.

Graduation, Pcm, Class Xii- 96%, Class X- 10/10 Gpa