Platforms
Adopt
Trial
-
24. Databricks Unity Catalog
Databricks Unity Catalog is a data governance solution for assets such as files, tables or machine learning models in a lakehouse. It's a managed version of the open-source Unity Catalog that can be used to govern and query data kept in external stores or under Databricks management. In the past our teams have worked with a variety of data management solutions such as Hive metastore or Microsoft Purview. However, Unity Catalog's combined support for governance, metastore management and data discovery makes it attractive because it reduces the need to manage multiple tools. One complication our team discovered is the lack of automatic disaster recovery in the Databricks-managed Unity Catalog. They were able to configure their own backup and restore functionality but a Databricks-provided solution would have been more convenient. Note that even though these governance platforms usually implement a centralized solution to ensure consistency across workspaces and workloads, the responsibility to govern can still be federated by enabling individual teams to govern their own assets.
-
25. FastChat
FastChat is an open platform for training, serving and evaluating large language models. Our teams use its model-serving capabilities to host multiple models — Llama 3.1 (8B and 70B), Mistral 7B and Llama-SQL — for different purposes, all in a consistent OpenAI API format. FastChat operates on a controller-worker architecture, allowing multiple workers to host different models. It supports worker types such as vLLM, LiteLLM and MLX. We use vLLM model workers for their high throughput capabilities. Depending on the use case — latency or throughput — different types of FastChat model workers can be created and scaled. For example, the model used for code suggestions in developer IDEs requires low latency and can be scaled with multiple FastChat workers to handle concurrent requests efficiently. In contrast, the model used for Text-to-SQL doesn't need multiple workers due to lower demand or different performance requirements. Our teams leverage FastChat's scaling capabilities for A/B testing. We configure FastChat workers with the same model but different hyperparameter values and pose identical questions to each, identifying optimal hyperparameter values. When transitioning models in live services, we conduct A/B tests to ensure seamless migration. For example, we recently migrated from CodeLlama 70B to Llama 3.1 70B for code suggestions. By running both models concurrently and comparing outputs, we verified the new model met or exceeded the previous model's performance without disrupting the developer experience.
-
26. GCP Vertex AI Agent Builder
GCP Vertex AI Agent Builder provides a flexible platform for creating AI agents using natural language or a code-first approach. The tool seamlessly integrates with enterprise data through third-party connectors and has all the necessary tools to build, prototype and deploy AI agents. As the need for AI agents grows, many teams struggle with understanding their benefits and implementation. GCP Vertex AI Agent Builder makes it easier for developers to prototype agents quickly and handle complex data tasks with minimal setup. Our developers have found it particularly useful for building agent-based systems — such as knowledge bases or automated support systems — that efficiently manage both structured and unstructured data. That makes it a valuable tool for developing AI-driven solutions.
-
27. Langfuse
LLMs function as black boxes, making it difficult to determine their behavior. Observability is crucial for opening this black box and understanding how LLM applications operate in production. Our teams have had positive experiences with Langfuse for observing, monitoring and evaluating LLM-based applications. Its tracing, analytics and evaluation capabilities allow us to analyze completion performance and accuracy, manage costs and latency and understand production usage patterns, thus facilitating continuous, data-driven improvements. Instrumentation data provides complete traceability of the request-response flow and intermediate steps, which can be used as test data to validate the application before deploying new changes. We've utilized Langfuse with RAG (retrieval-augmented generation), among other LLM architectures, and LLM-powered autonomous agents. In a RAG-based application, for example, analyzing low-scoring conversation traces helps identify which parts of the architecture — pre-retrieval, retrieval or generation — need refinement. Another option worth considering in this space is Langsmith.
-
28. Qdrant
Qdrant is an open-source vector similarity search engine and database written in Rust. It supports a wide range of text and multimodal dense vector embedding models. Our teams have used open-source embeddings like MiniLM-v6 and BGE for multiple product knowledge bases. We use Qdrant as an enterprise vector store with multi-tenancy to store vector embeddings as separate collections, isolating each product's knowledge base in storage. User access policies are managed in the application layer.
-
29. Vespa
Vespa is an open-source search engine and big data processing platform. It's particularly well-suited for applications that require low latency and high throughput. Our teams like Vespa's ability to implement hybrid search using multiple retrieval techniques, to efficiently filter and sort many types of metadata, to implement multi-phased ranking, to index multiple vectors (e.g., for each chunk) per document without duplicating all the metadata into separately indexed documents and to retrieve data from multiple indexed fields at once.
Hold
Unable to find something you expected to see?
Each edition of the Radar features blips reflecting what we came across during the previous six months. We might have covered what you are looking for on a previous Radar already. We sometimes cull things just because there are too many to talk about. A blip might also be missing because the Radar reflects our experience, it is not based on a comprehensive market analysis.
