A few years ago, we were called in to create an internal AL/ML curriculum for the largest tech & consulting company in the world. What we learned during that engagement was that most Fortune 1000 CEOs were frustrated that they had seen impressive AI/ML demos but they were not seeing the demos translate to implementations of value that reach their customers and benefit the bottom line.
There are many reasons for why AL/ML has been slow to proliferate in the enterprise. However, one of the key reasons has been the lack of ready-to-use tools and libraries that are commonly used by enterprise developers. Whereas very impressive AI/ML demos were in Jupityer notebooks, the true value to the company is only realized when the work in Jupityer notebooks is implemented in real code in the most common enterprise development stacks. The big companies have been able to deploy beneficial AI but the majority of companies are still struggling.
We believe that LangChain.js will solve this problem for LLM-based applications and beyond. Typescript/Javascript exists in many different stacks. It has a large developer base and footprint. LangChain.js will make it possible for AI to transition from demos in Jupityer notebooks to products delivering value in front of customers.
LangChain.js will also enable some use cases to skip the Jupityer notebook “data scientist” stage altogether. The power of LangChain’s Chains combined with Prompt Tempate repositories and Tool repositories will make the composing of AI products as simple as the way we develop products today (today we simply pick a set of publicly available NPM libraries and use them in our products).
In the past, we have worked with Apache Spark and other related cloud-based SaaS AI/ML for enterprise solutions. They have always felt like a heavy lift for many. That is what makes us extremely excited about LangChain.js. It has abstracted most of the heavy lifting that it feels very lightweight and approachable for many developers. We are excited and looking forward to educating developers on LangChain and using it for our customer implementations.