Arjun Kirti Patel

Authored "Contextual Retrieval with Pinecone and Anthropic" - technical breakdown of Applied the technique to a real-world, novel problem: asking visual/audio/image questions of webinar slideshows Recorded live webinar with over 900 registrants and 300+ attendees, over 1k views on YouTube

Authored "The Practitioner's Guide to E5" - walkthrough of Pinecone's landmark Inference endpoint complete walkthrough of multilingual embedding models and using Pinecone created novel example demonstrating monolingual, crosslingual, and multilingual search

Created "Refine Retrieval Quality with Pinecone Rerank" - technical breakdown of how top result on Google for "refine with rerank", along with AI Overview from Google

Hosted "The Magic of Multilingual Search" webinar Demonstrated how Pinecone's multilingual search can be used to search across multiple languages first principles approach to multilingual embedding models and search

Produced RAGBrag interview series with CEOs using Pinecone Interviewed CEOs from companies like Pinecone, Anthropic, and Cohere Featured on Pinecone's blog

CLI application that conducts agentic RAG over a statistical knowledge database, to identify statistical tests and methods for a given question, create examples of test application, and explain statistical concepts Uses Llamaindex Workflows, Cohere (tool use), Pinecone, Typer(cli), and structured generation using Llama 2 deployed on Baseten to conduct specific actions given user queries Presented at AI Tinkerers Chicago November 2024 Talk

Web application that assists competitive Pokémon VGC players in querying Pokémon data and performing speed checks Uses LLMs to parse natural language queries into JSON dicts, then passes to functions for calculation Built using Hugging Face, Langchain, Supabase, Modal, and Streamlit

First place winner in Cohere Hackathon #3 2022 Takes user prompts in Streamlit interface, identifies artistic style using Cohere classification Returns optimized prompts for iterative development from finetuned Cohere generation endpoints

coauthored blog post on overview of research published at HCOMP 2023

coauthored blog post on experimental design and landscape of third party AI content detection tooling overviews results of custom internal benchmark on detecting AI generated content

created results and coauthored blog post on results from applying internal AI detection models to real-world curated dataset from aforementioned experiment followup from HCOMP study on large scale dataset, demonstrating several-fold improvement in detection accuracy using behavioral techniques