Enterprise Knowledge Search
Stardog needed to remove the specialized-skill barrier for knowledge-graph queries while improving model quality and keeping customer data under tighter privacy and security control than third-party LLM services allowed.
Reported impact
Stardog turned knowledge-graph access into a conversational, self-service experience, reduced manual model-serving overhead, sped iteration with zero-downtime deployments and gained enough early customer enthusiasm to expand Voicebox beyond the initial access program.
Stardog used Databricks to add a conversational AI layer to its enterprise knowledge graph platform through a product called Voicebox. Instead of requiring users to master specialized query languages, Voicebox accepts natural-language questions, identifies the core concepts in each request and maps them to the structured concepts inside Stardog's semantic layer. Databricks Training is used to fine-tune task-specific models for this workload, giving Stardog tighter control over quality, privacy and security than the company found with off-the-shelf or third-party models. The resulting query flow combines the fine-tuned model with a vector-database lookup that aligns user language to enterprise concepts before returning the right knowledge-graph result in real time. Databricks version management and zero-downtime deployment further reduce manual model-serving overhead, so Stardog can iterate faster as it expands Voicebox from early-access customers to the broader cloud user base.
Users submit natural-language prompts to Voicebox, which uses a Databricks-trained model to identify core concepts in the request. Those concepts are aligned with structured enterprise data through the knowledge graph and supporting vector lookup, and the resulting answer is returned to the user in real time.
Databricks' version management and zero-downtime deployment capabilities have allowed Stardog to iterate and deploy new versions quickly and effectively.
That's why Stardog turned to Databricks to develop Voicebox.
The large language model (LLM) first identifies the core concepts in a user's natural language query and then sends it to a vector database to align those terms with actual concepts in the structured data sources.
Using large language models fine-tuned with Databricks, Voicebox acts as a conversational AI semantic layer that better connects users with their data through simple natural language prompts.
We quickly learned that if you have a smaller model that you fine-tune for a specific task, you can match or exceed the quality of OpenAI, and you have more control over data security and privacy.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
"streetAddress": "2101 Wilson Blvd 800", "addressLocality": "Arlington", "addressRegion": "VA", "postalCode": "22201", "addressCountry": "United States"
Related
Databricks Sample Jul 21, 2026 Zapier Zapier needed a better way to support AI-powered self-service analytics, conversational data access, and product personalization because legacy data silos and reliance on a central insights team meant many questions… 2–3 hours delivering ai-powered product and business insights Agent Bricks · Databricks Data + AI Platform · MLflow Databricks Sample Apr 2, 2026 nCino nCino needed to combine 13 years of banking data and trillions in loan history into compliant, domain-driven banking AI while keeping sensitive financial information under tight control and avoiding generic models that… Databricks AI/BI Genie · Databricks Data Intelligence Platform · Unity Catalog Databricks Sample Apr 1, 2026 Workday Workday needed a governed, AI-ready data foundation that could serve every line of business — from product telemetry and sales analytics to AI-driven marketing — and let technical and non-technical teams alike build… Every business unit launched sandbox workspaces in 6 months Databricks Data Intelligence Platform · Unity Catalog 