Jan 2025 – Jul 2025 · 6 mos · Virginia, United States · Remote

• Developed and implemented a RAG system for healthcare documents utilizing intelligent chunking algorithms tailored for medical billing data.
• Created an AI-driven feedback system by integrating OpenAI APIs with PostgreSQL, enhancing the scalability of healthcare platforms.
• Collaborated with cross-functional teams to ensure seamless integration of AI technologies in healthcare solutions.

Aug 2024 – Jan 2025 · 5 mos

• Co-developed QuOTE, to improve RAG performance by 20.7% using synthetic question generation for document embeddings
• Validated performance across 5 embedding models, proving effectiveness while maintaining faster queries than alternatives

Jan 2024 – Jul 2024 · 6 mos · On-site

• Contributed to the conception, design, and refinement of a unique LLM-based chatbot for The Washington Post, ensuring seamless integration with existing data and enhancing user engagement and interactivity.
• Developed a novel query time-scope based scoring model to improve RAG (Retrieval Augmented Generation) performance, and enhance the relevance of responses provided by the chatbot.
• Curated an original information retrieval based question-answer dataset to conduct thorough analysis and experimentation to fine-tune the scoring model

Jun 2022 – Jul 2022 · 1 mo · Gurugram, Haryana, India

• Implemented MLflow platform to manage ML Operations including experimentation, deployment, model registry etc.
• Trained ML models like decision tree, KNN etc, tracked their features on MLflow and served them using REST APIs
• Used Docker to containerize the entire project and deployed the container on Amazon EC2

Jul 2021 – Aug 2021 · 1 mo · Hyderabad, Telangana, India

• Applied zero shot learning for text mining applications (Resume Matching) in the absence of annotated data
• Used transformer based NLP models like BART to achieve the task of ranking resumes based on various labels
• Compared the effectiveness of language models like GPT3, BART and BERT in performing unsupervised NLP tasks

Apr 2021 – Jul 2021 · 3 mos · Hong Kong SAR

• Worked under the guidance of Prof. Dimitris Chatzopoulos - Computer Science, HKUST
• Experimented with state-of-the-art CNNs namely ResNet50, DenseNet169, MobileNet, VGG16 and EfficientNetB0 and evaluated their accuracy and training time, after introducing ‘early-exits’
• Carried out the experiments on two separate GPUs, trained the models for a total of 100 epochs and tabulated the results periodically, after every 20 epochs
• Investigated various strategies to minimize resource needs without compromising the accuracy of the model

Dec 2020 – Jan 2021 · 1 mo · Madhya Pradesh, India

• Designed, developed and deployed the conversational AI components of the Sarasvi Health AI Platform
• Planned and implemented the conversational AI assistant using advanced NLP and machine learning techniques