Yajat S

Architected and shipped end-to-end product features across services, frontend applications, and infrastructure, taking systems from zero to production. Designed scalable APIs, data models, and service boundaries to support evolving product requirements in a high-ambiguity environment. Led feature engineering cycles from ideation to deployment, making trade-offs between performance, scalability, and delivery speed. Implemented core architectural foundations including authentication, data persistence, caching, and observability, establishing early engineering standards. Built and iterated on user-facing workflows with modern frontend frameworks, translating product vision into performant, maintainable UI flows. Collaborated directly with founders on technical strategy, roadmap prioritisation, and system design decisions

Improved claims processing workflows by fine-tuning large language models using parameter-efficient techniques (LoRA, PEFT) on high-performance GPU infrastructure. Built internal AI tooling for automated data generation, model evaluation, and local inference pipelines leveraging the Hugging Face ecosystem.

Developed custom pipelines to transform unstructured data(pdfs,ppts) into structured formats for large language model ingestions. Led learning content conversion projects, engineering parsing workflows using an ensemble of layout analysis and vision models based on client specific format . Built Assessment generation engine and automated authoring ingestion pipelines, enhancing content validation, rubric mapping, and authoring platform integration.

Developed custom data ingestion pipelines for extracting unstructured data into large language models. Integrated AI based translation engines into the existing workflow, automating 8 hours per project. Built AI based closed caption generation and editing pipelines for internal product.

Developed a blockchain based supply chain traceability application catered towards the organisation’s supply chain transparency. Application was built with NextJS, MongoDB and Polygon Blockchain. Deployed the application on AWS and integrated rate limiting and CDN.

Engineered a Machine-Learning Model to Recommend foods for Diabetic population based on recent food data from the organisation. Developed a meal planner application and a food habit survey portal for Diabetic population using ReactJs,NodeJS and MySql. Built an android version of the Meal Planner using Flutter and dart.