Real-time Analytics
Before partnering with Databricks, Vannevar Labs struggled with the limitations of commercial models, such as GPT-4, which delivered suboptimal accuracy and was not cost-effective, especially given the multilingual complexities of their data. However, after using Databricks to integrate a compound AI system, the team successfully fine-tuned and deployed a highly accurate sentiment analysis model within just two weeks, reducing latency by 75% and achieving significant cost savings.
Reported impact
Vannevar Labs achieved measurable business impact on Databricks, including 2 weeks to deploy a fully trained sentiment analysis model; 76% model accuracy, compared with 64% using GPT-4; 75% reduction in model latency.
Defense-tech startup Vannevar Labs plays a critical role in supporting America’s strategic efforts to deter and de-escalate global conflicts. Operating at the forefront of defense technology, the company leverages advanced software and hardware to support hundreds of mission-focused users across various branches of the U.S. Department of Defense (DoD). Before partnering with Databricks, Vannevar Labs struggled with the limitations of commercial models, such as GPT-4, which delivered suboptimal accuracy and was not cost-effective, especially given the multilingual complexities of their data. However, after using Databricks to integrate a compound AI system, the team successfully fine-tuned and deployed a highly accurate sentiment analysis model within just two weeks, reducing latency by 75% and achieving significant cost savings.
Operational, customer, and transactional data lands in Databricks and is transformed into curated datasets
2 weeks to deploy a fully trained sentiment analysis model
75% reduction in model latency
76% model accuracy, compared with 64% using GPT-4
Initially, Punma’s team struggled with the limitations of using GPT-4 and prompt engineering, which did not yield satisfactory results for their specific needs.
“The best results we could get were around 65% accuracy, and it was overall too expensive for us,” Punma added.
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.
Related
Databricks Sample Jul 28, 2026 OpenAI OpenAI’s data platform powers one of the fastest-scaling products in tech history. ChatGPT, the OpenAI API, and the internal tools used by thousands of OpenAI employees require a foundation that can withstand sudden… $400K/month Storage costs eliminated by marketing team Databricks Data Warehouses · Databricks Jobs · Databricks SQL 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 Sep 22, 2025 Sovrn To operate at this scale, Sovrn needed to turn siloed Commerce and Ad Exchange data into actionable insights. Previously, fragmented systems, costly infrastructure and complex governance slowed innovation and increased… 30% Increase in publisher yield with smarter audience targeting and traffic shaping AI/BI · Delta Lake · Delta Sharing 