Unlock the Potential of Open Source AI with Llama

Health and Science

At the City of Hope National Medical Center, researchers are harnessing LLMs to develop tools for cancer doctors. One such tool matches cancer patients with relevant clinical trials. Medicine stands out as one of the most urgent applications for LLMs due to the scale and sensitivity of text data in clinics. As stated by Kingson Man, Principal Data Scientist at City of Hope, 40% of cancer clinical trials fail due to the inability to find enough eligible patients. Only 3-7% of patients actually participate in trials. To bridge this gap, City of Hope constructed a matching tool that analyzes whether specific clinical trial eligibility criteria apply to their current patients and generates a matching score. With this score, doctors can uncover additional, perhaps overlooked clinical trials for their patients.

OtonoCo utilizes open source LLMs to power Gene Outlook, a platform aimed at accelerating the discovery and extraction of insights from data, particularly in genomics. Researchers use Gene Outlook to identify important gene signatures or other gene-related data in cancer patients. Then, Llama transforms the gene symbol into content that scientists and businesses can easily comprehend. OtonCo has also developed an organism-aware LLM on top of Llama called GeneTuned LLM and is in the process of expanding Gene Outlook to areas like gut health and microbiome. As Jong Hang Siong, Founder and Chairman of OtonoCo, puts it, "With open source, we can focus solely on the problem and work backward to determine the suitable open source technology."

Education

DLYog Lab was co-founded by Tarun Chawdhury and Mousumi Chawdhury, inspired by their son's special needs. They developed a GenAI-powered app that simplifies the creation of Individualized Education Plans (IEPs) for special education. The app processes transcripts from IEP meetings involving parents, the child, teachers, and specialists to generate an education plan detailing goals and services for the child. Tarun Chawdhury, Co-founder of DLYog Lab, emphasizes that Llama enables small players like them to fulfill their social good dreams. It was initially created for his son's benefit but can now be extended to many other parents.

DLYog's app allows teachers to quickly create personalized plans based on a child's profile. Llama's publicly accessible open source architecture can be easily scaled for small institutions like private schools. Since the app can be deployed in a private data center with minimal infrastructure requirements, small schools and private institutions with data security concerns can adopt it within their own premises. With the latest version of Llama, DLYog Lab's app now analyzes nonverbal cues such as facial expressions and eye movements, providing insights beyond traditional IEP approaches.

Pratham Education Foundation (PEF) in India used Llama to create a solution for young mothers to learn about childcare and early childhood education. Accessing this information on the open web can be challenging as it often provides confusing or conflicting data. PEF's WhatsApp-based chatbot spans 12 Indian languages, and mothers can interact with it through voice instead of typing. The bot generates audio and video answers based on verified documents and also sends a video link for easier information digestion. Over 40,000 mothers in India have used this product. According to Nishant Baghel, Director of Technology Innovation at PEF, Llama serves as PEF's "backbone" for generating and consuming knowledge.

Small Business and Entrepreneurship

HiiiWAV, an incubator in Oakland, California, teaches artists, especially Black musicians, how to build startups using AI. Their goal is to help these artists recognize and utilize the power of AI and machine learning. Since early-stage startups need to keep costs low, HiiiWAV recommends open source technology, which equalizes opportunities by providing access to often expensive and highly technical tools. Among the available options, they consider Llama to be "top of the list" as it not only reduces costs but also enables them to build a community and share information freely. Bosko Kante, Executive Director of HiiiWAV, states, "Off-the-shelf closed-source options become expensive with usage and are difficult to build with. We are also developing AI-powered hardware, so we need the control and freedom that is missing from closed source options. Open source is clearly the way forward."

Entrepreneurs in the HiiiWAV incubator have used open source AI to create various tools. GooRoo is a mobile voice-assisted record producer, and Choice Scores is an AI-powered platform that converts hip-hop songs into musical scores. AIIRA, the AI Institute for Resilient Agriculture, builds AI tools to assist agriculturists with their daily operations. Their aim is to encourage AI adoption among farmers to help them overcome challenges in crop improvement and production. Llama's open source technology allows for multilingual, geographic, and context-dependent fine-tuning, enabling the organization to have different models suitable for specific regions. AIIRA uses Llama to generate personalized recommendations for farmers to address pests and weeds. Chinmay Hegde, an AIIRA member and professor at New York University, says, "Open source AI's ability to scale is transformative. It works in the United States, Africa, India, and across the globe. I hope this open source spirit continues in the future to enable us to leverage the latest AI advancements."