Our secure integrations now cover a majority of the world’s most popular LLM providers, including Microsoft, Google and Amazon.
Today Sama announced new model integration offerings for some of the world’s most popular LLMs as part of our Sama Gen AI solution. With these developments, Sama can seamlessly access and retrieve data from top open-source GenAI models and LLMs, then send it back through the same protected pipeline. This not only reduces the risk of data exposure, but makes it easier to add a human-in-the-loop (HITL) component to the model lifecycle — crucial for more responsible AI development and deployment.
“Generative AI is at a pivotal point in its development. We are now wrestling with larger questions about how to develop models more responsibly without sacrificing quality or efficiency. Our new model integrations neatly solve this issue by ensuring that data is transferred directly to our dedicated workforce in secure facilities,” said Duncan Curtis, SVP of AI product and technology at Sama.
“We can then easily send our annotations to the model platform or an on-prem model, depending on a client’s needs, again increasing security and improving efficiency. More responsibly developed AI is key to creating more ethical AI, and we’re proud to be playing a part in pushing this process forward.”
Sama’s platform can now integrate with a number of new model providers, including but not limited to the following: Mistral, Microsoft, Amazon and Databricks. These integrations join previous integrations with Cohere, Anthropic, OpenAI and Inflection among other companies. For all providers, Sama can evaluate both prompts and model responses, scoring and ranking them across a variety of client-defined dimensions, such as coherence. These integrations also support Retrieval-Augmented Generation (RAG) evaluation for improved accuracy and reliability of model output. Finally, direct data transfer accelerates supervised fine-tuning, allowing companies to build on these pre-trained LLMs to create models with knowledge bases specific to their industry or business with the help of Sama prompt engineering.
Sama plans to further expand its integration capabilities to other API-driven models in the near future. Its solutions are designed to scale to all project sizes, including some of the largest open-source models in the world, and can reduce project starting times by several weeks. The company’s HITL approach involves a vetted, experienced and diverse workforce of over 5,000 annotators providing critical feedback loops to models to validate they are behaving to the client’s specific parameters. This feedback occurs during the entire model development process, including data creation, supervised fine-tuning, LLM optimization and ongoing model evaluation. By consistently providing this feedback, the company helps clients develop their models in a more responsible way. Sama is itself compliant with key regulatory directives and can help its customers achieve similar compliance as a supplier. Furthermore, Sama’s focus on paying a living wage and providing benefits to its employees help promote a more ethical AI supply chain.
All of Sama’s services leverage SamaHub™, a collaborative workspace where clients and team members can directly communicate on workflows and complete reporting to track their project’s progress. Sama’s work is backed by SamaAssure™, the industry’s highest quality guarantee, which routinely delivers a 98% first batch acceptance rate. Projects leverage SamaIQ™, a combination of Human in the Loop assessments and proprietary algorithms, to proactively surface additional insights into a model’s vulnerabilities.
Sama has been named the Best AI Model Validation Solution by the annual AI Breakthrough Awards, which recognize the breakthrough companies, technologies, products and services in the field of AI around the world.
Large language models (LLMs) have emerged as powerful tools capable of generating human-like text, understanding complex queries, and performing a wide range of language-related tasks. Creating them from scratch however, can be costly and time consuming. Supervised fine-tuning has emerged as a way to take existing LLMs and hone them to a specific task or domain faster.
Sama provides ML Professionals and AI team Leads with an indispensable solution for Computer Vision data labeling.
