Giga ML

Launched in 2023, Indo-US firm Giga ML helps organizations secure on-premise deployment of AI and Large Language Model (LLM)-powered solutions for multiple internal and external use cases. 

The platform allows businesses to deploy LLMs as robust as GPT-4 directly on their own servers, eliminating the need to send sensitive information to external servers.

Giga ML’s X1 Large 32k model is a pre-trained and comprehensively fine-tuned iteration of Meta’s Llama2 70B 4k model. Organizations have the capability to take Giga ML’s base model and further continue pre-training and fine-tune it to meet their unique requirements. The organizations have the ability to pre-train with specific text data, enhancing customization for different sectors. It also provides flexible finetuning options for specific output structures, ensuring nuanced and relevant responses. Additionally, the model guarantees secure handling of sensitive data through on-premise solutions. The firm claims that other competitors have still not been able to fully fine-tune Llama 2 with 32k context length. Giga ML also has its own inference optimization algorithms, which it claims provide superior performance and reducing costs. It claims its innovative approach leads to a 70% reduction in costs compared to GPT-4 models and a 300% increase in output delivery speed. 

Key customers and partnerships

Giga ML’s offerings can be used by firms in various industries such as healthcare, legal, and finance. Giga ML sees enterprises adopt their platform for use cases like customer support, internal knowledge search, and code generation for the productivity of engineering teams. 

In August 2023, Giga ML partnered with Mano AI, an LLM-powered conversational search solution to unveil the on-premise Retriever Augmented Generation (RAG). This collaboration enables extensive, secure, and responsive document interaction by providing on-premise embeddings and vector databases for enhanced data privacy.