Ask Priya" is a pioneering AI chatbot designed to streamline the process of acquiring US immigration information. Leveraging data from the USCIS website, this bot addresses queries relating to American immigration with enhanced precision. Its backbone consists of the llama-index and Google Vertex AI, which significantly mitigate common issues associated with conventional large language models (LLMs) like ChatGPT. When prompted with a question, "Ask Priya" sifts through its USCIS-indexed data, selecting relevant documents to guide the LLM's response, thereby reducing 'hallucinations' – instances where models inaccurately fabricate plausible-sounding information. In light of the substantial international arrivals to the United States, which totaled 22.1 million in 2021, and the presence of over 1.08 million international students with active study visas as of January 2023, the demand for tailored immigration services is immense. Additionally, with 483,927 registrations for H1B visas in 2023 and only 127,600 selected to apply, the complexity and breadth of US immigration, encompassing 35 nonimmigrant and 13 immigrant visa categories, becomes evident. "Ask Priya," in its beta phase, is gearing up for an imminent rollout. The immediate agenda includes multilingual query handling for Trulens evaluations, integrating these evaluations with GitHub actions, transitioning from Streamlit Community Cloud to Google Cloud, and storing chat logs on Google Cloud Storage. Looking towards the future, the development team aims to refine retrieval methods for better context relevance scoring. Plans include enhancing Trulens evaluations through finetuning, in addition to the current use of Retrieval-Augmented Generation (RAG). Moreover, the team is focused on creating a persistent chat memory, enabling users to access their chat history, thereby offering a more seamless and personalized experience.
Category tags:"Glad to see that you've continued adding improvements to Ask Priya. Your roadmap for improvements shows careful thought and planning. I would encourage you to think about the monetization potential of what you're building."
Hardik Vala
Founder
"i remember your app from past hackathon, it really captured my attention. great work to improve the previous version of your chatbot. it is time to go for the marked and try to make full use if it."
Walaa Nasr Elghitany
Data scientist and doctor
"Loved the system design explanations and the improvements you made over the last version. Happy to see the product making progress in terms of the performance optimization decisions you took. Hope to see it develop into a monetized product in future."
Abhinaw Tiwari
AI focussed Software Engineer