Nova AI's approach is driven by the experiences of its founders, former Google and Meta engineers, who found that large enterprises were wary of their data being used by OpenAI. The company is also responding to the rapid progress of open-source models, which are proving to be more than sufficient for specific tasks such as writing tests. Nova AI's focus is on fine-tuning its models for this specific task, rather than building a broad-based model.
Key takeaways:
- Nova AI, a startup focused on code testing, has raised a $1 million pre-seed round and is targeting mid-size to large enterprises with complex code-bases.
- The company's technology uses GenAI to automatically build tests, particularly for continuous integration and continuous delivery/deployment (CI/CD) environments.
- Unlike many AI startups, Nova AI is minimizing its use of OpenAI’s GPT-4 due to trust issues from enterprises, and is instead relying on open source models like Llama and StarCoder, as well as building its own models.
- Open source AI models are proving to be more cost-effective and sufficient for targeted specific tasks, and are rapidly progressing, with new versions like Meta's Llama earning recognition in technology circles.