Excited to share the latest LinkedIn Eng blog, where Juan Pablo Bottaro and I pull back the curtain on our experience developing LinkedIn’s generative AI-powered products. The biggest takeaway for me? Response quality and latency are imperative when developing a seamless GenAI-powered experience. The team is continuing to apply these learnings to our products as we push to turn every feed and job posting into a springboard for opportunity for our members.
This is excellent and I really appreciate the transparency because it's so helpful to be able to compare your experiences to the small-co startup experience of working through many of the same challenges. { "type": "json_object" } being added to GPT-4 Turbo was a huge improvement for me!
Very insightful, lots of learning 👍🏽
Great work, and a very insightful read. Thanks for sharing!
Staff Software Engineer @ LinkedIn
3mo"a unified skill registry to dynamically discover and invoke APIs/agents packaged as LLM friendly skills" I am curious if this skill registry could grow to involve all internal APIs irrespective of if they are used for GenAI products. This could then be used by developers to ask LLM to generate code by just inputting business use cases. Thoughts?