Our AI R&D experience in EU-funded projects has enabled us to provide two kinds of LLM services:
LLM fine-tuning (LLAMA models and others):
This service helps you exploit the full potential of large language models for your specific use case (e.g., creating a chatbot to support your customers, or improving the interface of your robot).
The service covers the complete fine-tuning process:
- Model selection and architecture design: Our expert analysis tailors the ideal LLM (such as the LLAMA variants 7B, 13B, or 70B and other alternatives) to your specific use case, with custom architecture enhancements designed to boost model performance and unlock new capabilities in your domain
- Data preparation and augmentation: domain-specific datasets, data cleaning and preprocessing. Augmentation techniques and synthetic data generation when required.
- Fine-tuning/Optimization through different techniques: RAG, LoRA, QLoRA P-Tuning v2, etc.
- Model evaluation: Testing of the new optimized model and refinement.
Connecting LLMs and Robots: The say-can approach
The approach described in the paper “Do As I Can, Not As I Say: Grounding Language in Robotic Affordances” (see https://say-can.github.io/ ) bridges the gap between large language models (LLMs) and robotic systems by providing a way to connect the high-level semantic knowledge represented in LLMs to real-world tasks that can be executed by robots.
The eProsima team will help you implement this approach for your robotic project by defining the list of possible robot tasks and their feasibility, and translating the output of an LLM prompt into a list of these tasks to create a possible plan to execute the command.
Need more info: