Fine-tuned BERT and ALBERT with span-labelling heads to build an extractive question answering system that extracts answers as a contiguous span of text from a passage. A few of its applications are customer service, knowledge management, legal and financial information extraction, and public information system.

Tech: Python, Data Cleaning, Fine-tuning, Transfer Learning

Management: team meetings, presentation, documentation

Developed a Generative AI–powered mobile application to train nurses for Occupational English Test (OET) preparation. It consisted of a RAG-based chatbot to help nurses prepare for the test by comparing their responses to questions with AI-based responses to same questions, improving self-assessment and training efficiency.

Tech: LlamaIndex, RAG, VectorDB, React Native, Express.js, Python, MongoDB, FastAPI, Microsoft Azure

Management: project management, vendor management, software requirements specifications

Designed, developed, and managed two frontends and a backend for a children's quizzing website for MyGov.in, aimed at improving the government's digital engagement with children.

Tech: ReactJs, NestJS, PostgreSQL, Redis, Docker, Grafana, Microsoft Azure, software architecture, database design, microservices

Management: Scrum, JIRA, project management, client meetings, software requirements specifications

Built a scalable backend system for motor and health insurance renewals for an insurance aggregator, enabling seamless policy renewals, validations, and integrations with external partners.

Tech: NestJS, PostgreSQL, MongoDB, Amazon Web Services, microservices, RabbitMQ

Management: Scrum, JIRA, project management, client meetings

Developed an automated scheduling system that calculates optimal travel routes and heuristically allocates hospital operation theatre time slots to surgeons, avoiding scheduling conflicts for surgeons traveling across multiple hospitals in a day.

Tech: Python, data cleaning, vehicle routing (TSP-Dijkstra’s Algorithm), geocoding, heuristic, MongoDB, Django

Built a hybrid recommendation system combining content-filtering and collaborative-filtering techniques to deliver personalized apparel recommendations, improving relevance, customer engagement, and sales.

Tech: R, exploratory data analysis, data preprocessing, feature engineering, content filtering, collaborative filtering, hierarchical clustering, CART, random forest, time weighting, SQL, Flask

Built a predictive analytics model to anticipate valve trip events in gas turbines using machine learning and deep learning, enabling proactive maintenance and reducing unplanned downtime, and thus preventing critical damage to turbines.

Tech: Python, SQL, exploratory data analysis, data preprocessing, machine learning, deep learning, data engineering, feature engineering, information fusion