Proxy-FDA: Proxy-Based Feature Distribution Alignment for Fine-Tuning Vision Foundation Models Without Forgetting

Vision foundation models pre-trained on massive data encode rich representations of real-world concepts, which can be adapted to downstream tasks by fine-tuning. However, fine-tuning foundation models on one task often leads to the issue of concept forgetting on other tasks. Recent methods of robust fine-tuning aim to mitigate forgetting of prior knowledge without affecting the …

ml 18410 arch diag

Impel enhances automotive dealership customer experience with fine-tuned LLMs on Amazon SageMaker

This post is co-written with Tatia Tsmindashvili, Ana Kolkhidashvili, Guram Dentoshvili, Dachi Choladze from Impel. Impel transforms automotive retail through an AI-powered customer lifecycle management solution that drives dealership operations and customer interactions. Their core product, Sales AI, provides all-day personalized customer engagement, handling vehicle-specific questions and automotive trade-in and financing inquiries. By replacing their …

maxresdefault

Announcing new MCP integrations to Google Cloud Databases to enable AI-assisted development

Last month at Google Cloud Next ‘25, we announced MCP Toolbox for Databases to make it easier to connect generative AI agents to databases, and automate core enterprise workflows. MCP Toolbox for Databases (Toolbox) is an open-source Model Context Protocol (MCP) server that allows developers to easily connect gen AI agents to enterprise data. It …

Analyzing the Effect of Linguistic Similarity on Cross-Lingual Transfer: Tasks and Input Representations Matter

Cross-lingual transfer is a popular approach to increase the amount of training data for NLP tasks in a low-resource context. However, the best strategy to decide which cross-lingual data to include is unclear. Prior research often focuses on a small set of languages from a few language families or a single task. It is still …

ML 18605 mcp architecture

Unlocking the power of Model Context Protocol (MCP) on AWS

We’ve witnessed remarkable advances in model capabilities as generative AI companies have invested in developing their offerings. Language models such as Anthropic’s Claude Opus 4 & Sonnet 4 and Amazon Nova on Amazon Bedrock can reason, write, and generate responses with increasing sophistication. But even as these models grow more powerful, they can only work …

1 Qls8qpSmax 1000x1000 1

Emulating the air-gapped experience: GDC Sandbox is now generally available

Many organizations in regulated industries and the public sector that want to start using generative AI face significant challenges in adopting cloud-based AI solutions due to stringent regulatory mandates, sovereignty requirements, the need for low-latency processing, and the sheer scale of their on-premises data. Together, these can all present institutional blockers to AI adoption, and …

Prompting Whisper for Improved Verbatim Transcription and End-to-end Miscue Detection

*Equal Contributors Identifying mistakes (i.e., miscues) made while reading aloud is commonly approached post-hoc by comparing automatic speech recognition (ASR) transcriptions to the target reading text. However, post-hoc methods perform poorly when ASR inaccurately transcribes verbatim speech. To improve on current methods for reading error annotation, we propose a novel end-to-end architecture that incorporates the …

ML 18340 001 graph constructed

Build GraphRAG applications using Amazon Bedrock Knowledge Bases

In these days, it is more common to companies adopting AI-first strategy to stay competitive and more efficient. As generative AI adoption grows, the technology’s ability to solve problems is also improving (an example is the use case to generate comprehensive market report). One way to simplify the growing complexity of problems to be solved …