Sep 2024 – Present · 1 yr 8 mos · San Diego, California, United States

• Developed Claire, the TRAI Chatbot for sepsis prediction, using Flutter and ML pipelines on historical EHR data, currently being trialed across three departments at UCSD hospitals.
• Built a context- and knowledge-driven conversational agent with embedded medical ontologies to conduct comprehensive patient history intake.
• Containerized microservices with Docker, automated CI/CD pipelines via GitHub Actions, and deployed on AWS EC2 for scalable and secure operations.
• Created a FastAPI‑served LangChain NLP framework to extract and synthesize treatments, medications, and clinical history from chat transcripts.
• Integrated Epic MyChart workflows via FHIR APIs, enabling seamless import of patient encounters into the app workflow to enrich the agent’s diagnostic context.

Jun 2024 – Sep 2024 · 3 mos · Maharashtra, India · Hybrid

• Architected a Multitenant RESTful API: Designed and implemented a versatile multitenant RESTful API to power the participant management portal, ensuring efficient backend functionality.
• Crafted a User-Centric Frontend with React: Developed a responsive frontend interface using React, enhancing user navigation and enabling seamless access to their data.
• Engineered Customized Flutter Applications: Designed and developed tailored Flutter applications for diverse clients, optimizing event and participant management processes for enhanced efficiency.
• Created a Volunteer Communication Mobile App: Developed a dedicated mobile application for volunteer communication, seamlessly integrating with 100ms Audio rooms and simplifying participant check-ins.
• Designed a Personalized Templating Engine: Built a comprehensive templating engine using LiquidJS, facilitating personalized participant communication through customizable templates.
• Integrated a Streamlined Payment Gateway: Implemented a seamless payment gateway by integrating with PhonePe, reducing dependency on external third-party services for enhanced transaction efficiency.
• In Progress: Photo Tagging Software Development: Currently engaged in the design and development of advanced photo tagging software, categorizing participant photos based on bib numbers for improved organization.
• Implemented Secure Authentication Management: Implemented robust authentication across the platform using Zitadel, a secure authentication management application, ensuring data security and user privacy.

Aug 2023 – May 2024 · 9 mos · Pune, Maharashtra, India · Remote

• Contributed to Dataset Generation for Low-Resource Indic Languages: Spearheaded the creation of datasets tailored for low-resource Indic languages to facilitate question-answering tasks, culminating in a publication at ICON 23, showcasing dedication to linguistic research and cross-cultural communication.
• Engineered a Robust Approach for Answer Span Calculation: Devised a rigorous methodology utilizing dynamic programming to accurately determine answer spans, while concurrently establishing an efficient pipeline for data generation, demonstrating proficiency in algorithmic design and data processing.
• Led Research Group on Automatic Assessment of Multi-Modal Answer Sheets: Investigated innovative techniques for the automatic evaluation of multi-modal answer sheets within the STEM domain, with a specific focus on evaluating textual and diagrammatic content using advanced Graph Neural Networks (GNN), CRAFT and Yolo V5, showcasing expertise in interdisciplinary research and machine learning methodologies.

Jun 2023 – Jul 2023 · 1 mo · Bengaluru, Karnataka, India · On-site

• Developed Advanced Fuzzers: Engineered sophisticated fuzzers specifically targeting the Android Bluetooth stack, leveraging the power of LLVM to ensure robust testing.
• Enhanced Code Coverage: Focused on significantly improving the code coverage of the Bluetooth stack, ensuring comprehensive testing and identification of potential vulnerabilities.
• Utilized LLVM Toolset: Employed LLVM’s extensive toolset to create highly effective fuzzing strategies, optimizing the identification of bugs and security flaws within the Bluetooth implementation.
• Implemented Test Strategies: Designed and executed test strategies that maximized the effectiveness of fuzzing efforts, contributing to the stability and security of the Android Bluetooth stack.
• Collaborated with Development Teams: Worked closely with cross-functional teams to integrate fuzzing tools and methodologies into the development pipeline, ensuring continuous improvement in code quality and security.

Jan 2023 – Jun 2024 · 1 yr 5 mos · Pune, Maharashtra, India · Hybrid

• Developed a Versatile AI Inference Engine: Engineered a robust AI Inference engine, built using LangChain and FastAPI, capable of integrating any AI models as needed. This automated the process of creating observational data from files to the platform, generating personalized care plans, and extracting key patient insights from data such as survey responses and observational files. This was enabled by a Template engine to store templates and prompts which was built over AWS SAM using Lambda functions that ensured efficient storage and retrieval of necessary information.
• Designed Template Engine: Built a template engine for storing use cases, templates, and prompts, enabling the AI inference engine. Utilized AWS SAM with Lambda for efficient information storage and retrieval.
• Implemented Event Worker Module: Created an event worker module to execute tasks in response to events. Utilized RabbitMQ for message queuing and a CRON job for task execution based on configurations.
• Built Multitenant Notification Service: Developed a notification service sending alerts via FCM, email, and text. Included user notification preferences for customized delivery.
• Managed Health Plans: Implemented personalized health plan management for patients, ensuring comprehensive care.
• Enhanced Templating Engine: Improved a templating engine with LiquidJS for personalized messaging, boosting user engagement.
• Streamlined PDF Generation: Developed PDF generation from HTML using wkhtmltopdf and integrated delivery via Nodemailer.
• Contributed to Video Conferencing Platform: Assisted in developing a video conferencing platform using 100MS, focusing on API development.
• Refined Platform APIs: Contributed to the platform's overall development, particularly in designing and refining APIs for EHR integration.
• Optimized Deployment and Cost Efficiency: Reduced Docker container sizes, lowering AWS ECS deployment costs. Established CI/CD pipelines for efficient deployments.

Oct 2022 – Jul 2023 · 9 mos · Hybrid

• Conducted In-Depth Research on Text Synthesis and Analysis: Undertook extensive research by reviewing and synthesizing information from numerous academic papers, enhancing the understanding and application of advanced text synthesis and analysis techniques.
• Implemented Word2Vec Model for Transcript Generation: Developed a Word2Vec model to accurately generate transcripts from audio conversations, improving the efficiency and accuracy of speech-to-text processes.
• Integrated Advanced NLP Models for Entity Extraction and NER: Utilized Stanford CoreNLP and Flair NLP models to perform sophisticated audio entity extraction and Named Entity Recognition (NER), enhancing the capability of the system to identify and categorize key information from audio data.

Sep 2022 – Dec 2022 · 3 mos · Kolhapur, Maharashtra, India · On-site

• Architected and Deployed a Scalable Postgres Database: Leveraged Prisma to design and implement a robust Postgres database, ensuring seamless data management and high performance. Successfully deployed the database on Render, optimizing for reliability and scalability.
• Engineered a High-Performance Backend API: Developed a comprehensive RESTful API in TypeScript, focusing on performance, security, and maintainability. Deployed the API on Render, facilitating efficient and scalable backend operations.
• Crafted a Dynamic Flutter Application: Designed and implemented a feature-rich Flutter application, seamlessly integrating with the backend API. Ensured a smooth and responsive user experience through effective API integration and performance optimization.