We are thrilled to share our work on Pulsar Agent, our embedded AI assistant within the Pulsar for Salesforce application. We think this is truly groundbreaking work and a glimpse into what’s ahead.

With the rapid evolution of AI, the announcement of the DeepSeek and Stanford S1 models in late 2024 truly changed the game. At Luminix, we recognized these developments validated the incredible opportunity we saw to bring AI advancements to enterprise companies—without the massive financial burden typically associated with cutting-edge AI. What if you could integrate a highly specialized AI model directly into your enterprise mobile app, one that handles heavy lifting and drives productivity for users? That’s exactly what we are doing with Pulsar for Salesforce.

Already a robust, local-first app capable of managing millions of offline records, Pulsar now includes an embedded AI model designed to assist users by performing tasks, offering insights, and delivering tailored support. This marks the beginning of what we see as the future of applications: most will eventually incorporate embedded language models (ELMs) that become smaller, more efficient, and more specialized. We believe that targeted AI models, like the one we’ve embedded, are the way forward.

For more insights, check out my full thoughts here: Local-First AI Native Models Can Transform the Enterprise.

In today’s demo, we’re showcasing a first glimpse of what an embedded AI model in a mobile app can achieve—and this is just the beginning. In the recorded demo, you’ll see the AI Pulsar Agent, assist field reps by transcribing user commands, offering relevant knowledge articles, and even providing predictive analysis. All of this is powered by local data and dynamic, on-the-fly fine-tuning.

While we’re excited about this development, our customers are equally enthusiastic to give it a try. You may notice that we’re conducting the demo on a Windows machine. This is because optimal choices need to be made regarding size, performance, and accuracy while ensuring the resulting binary remains manageable. We continue to monitor how platforms like Google Play and the iOS App Store evolve to handle larger app footprints. For now, Windows offers the most viable option for delivering larger app binaries that can be easily installed by corporate IT departments.

Stay tuned for even more exciting updates as we expand to additional platforms!