- Insituate's Newsletter
- Posts
- Product Update #2
Product Update #2
Data store; Knowledge base; Finetuning is all you need.
Hello,
We’re very excited to share the second update with you! This one is special since we have now become a part of an exclusive community of 37 hottest GenAI startups in India. 🎉
Thank you for subscribing to the Insituate’s newsletter. You’re now part of a community that’s shaping the future of AI and innovation.
Product Updates
There are two facets when it comes to enterprise data - security & accessibility. The data should best stay with the business with proper RBAC and ABAC policies in place. If the user wishes to put in new data, he/she should have access to multiple sources of data ingestion i.e. SQL, Amazon S3, local directory, Google Drive, etc.
We present two frameworks that fulfil these conditions, all the more providing the best user experience there is.
• Knowledge Base - Along with our lighthouse partners in enterprises we developed a pipeline to connect the copilot to your data securely. Ground your API on the data, giving the ability to produce responsible, well-cited responses. The copilot uses advanced retrieval and search to best utilise the uploaded information.
Domain Knowledge tab on the Insituate platform
• Data Store - Now, every business has multiple departments and thus requires multiple copilots for the professionals to use as tools so they can focus on high-value work. This has led to strict policies on the platform for who can access data that is uploaded to the platform but is still with the organization. The Data Store lets you find all the data in one place if given access, so that you may not need to repeatedly update the data in each copilot separately. You can observe and improve with a single click!
Data Store on the Insituate platform
• Finetuning - More than a buzzword. Let’s say that you want to get the same performance out of a 7B model that you expect from the biggest commercial models out there such as GPT-4o. And you want to optimise for cost. Plus you don’t want to keep the same 10,000 tokens of domain knowledge in your model’s context window and delay the responses at the very least. What do you do?
You finetune!
The fine-tuning section on the Insituate platform
Spotlights
We are beyond thrilled to announce that Insituate has been handpicked as one of the 37 trailblazing startups for Nasscom’s GenAI Foundry program's second cohort, standing out among 130+ fierce contenders! This selection is a testament to our relentless drive to empower businesses to build state-of-the-art, secure copilots on-site, within a day.
Technology
Check out our blog on AutoLLM.
Whether fine-tuning is the right call for you or not is an important decision that has to be considered at all iterations of your GenAI dev cycle. You can prompt engineer or tweak your way through the hyperparameters of different models themselves. But will you do this for 10,000,000 possible configurations?
Need Help?
Reach out [email protected]
Get started here, or if you’d like to bring these features to your enterprise please get in touch.
— The Insituate Team