Nasir Shadravan

Extracting Actionable Insights from Customer Calls: A Practical Example

A lot of businesses do their customer support through phone calls. Usually, customers prefer to reach out for the phone to get their issues solved rather than to get into a ping pong email interaction to solve a problem. It is faster, interactive and more human. Customer support calls are direct link to the customer’s thoughts and needs. Businesses receive direct feedback and critisism into the operations and delivery of the service or product.

GPT4 Vision Structured Image Data

The OpenAI GPT-4 vision model allows you to understand images and generate text about them. This model is used by ChatGPT when you upload an image and ask a question about it. It is quite powerful and can be used in a variety of applications. The model is also accessible via the OpenAI API. You can upload an image and run a prompt against it to generate a text output.

Architecting LLM Powered Software

The first wave of LLM apps were famously looked down upon as “gpt wrappers”. Single prompt templates generating simple text. But if you’re building an AI first application, you need an AI design approach. And that’s the moat. You combine the engineering approaches to create a new generation of software. A software that is more natural and more human. Or can process human generated knowledge. There is a lot of hype around AI and Large Language Models.

About Me

My name is Nasir Shadravan. I’m located in Amsterdam. I’m an experienced software engineer with ~20 years of experience building software. I’ve worked on a wide range of projects, from small startups to large enterprises. From J2EE to Django. From microservices monoliths. I mostly write python. For the web side, I use HTMX, vanilla JS and Hyperscript. I avoid writing Javascript as much as possible. I prefer monolith apps. They are easier to reason about and develop.