How to Use the OpenAI API Effectively and Not Just Be a Meme Startup

DaherDaher
2 min read

It's quite simple:

You can either use a tool well or misuse it; there's no middle ground.

In 2023, I was working on a feature aimed at extracting information from CVs submitted by candidates applying for job postings.

We explored several companies that offered an API for this service. One of these companies charged $0.25 USD to extract information from each CV.

Parsing thousands of CVs meant that every 1000 CVs would cost us $250 USD!

While this amount might seem small, it can significantly increase the costs of a secondary feature.

Shortly after, OpenAI released some of its APIs, and our team implemented this feature at a much lower cost.

Also, in 2023, "AI Startups" began to emerge, offering services with OpenAI integration at their core.

By the end of 2023, with the launch of GPT extensions and access to new APIs, these "AI Startups" simply Disappear.

The issue isn't with using OPEN AI's APIs to create value!

The problem arises when the value you offer doesn't extend beyond a simple integration or relies on a third party that has more infrastructure than you and can develop it further, leaving you with nothing.

After these two stories, let's briefly analyse some of the APIs currently available:

  • Text to Speech:

    • The Audio API's speech endpoint, based on our TTS (text-to-speech) model, includes 6 built-in voices and can be used to narrate a written blog post, produce spoken audio in multiple languages, and provide real-time audio output using streaming.
  • Speech to Text:

    • The Audio API offers two speech-to-text endpoints, transcriptions and translations, based on our state-of-the-art open-source large-v2 Whisper model.

    • They can transcribe audio into the language it's in and translate and transcribe the audio into English.

  • Embedding:

    • OpenAI’s text embeddings measure text string relatedness, commonly used for search, clustering, recommendations, anomaly detection, diversity measurement, and classification.
  • Assistant (Beta):

    • The Assistants API allows building AI assistants within your applications, leveraging models, tools, and knowledge to respond to user queries.

    • It currently supports Code Interpreter, Retrieval, and Function calling tools, with plans to release more OpenAI-built tools and allow for your own tools on our platform.

We must understand these APIs for what they are: tools that can save us an infinite amount of work and should be used to enhance our product.

How you combine and integrate them into your work is only limited by your imagination.

Although there are still many limitations and they should not be used in critical environments, beginning to add them to our arsenal opens up a world of possibilities.

What interesting problem are you solving with these integrations?

0
Subscribe to my newsletter

Read articles from Daher directly inside your inbox. Subscribe to the newsletter, and don't miss out.

Written by

Daher
Daher

Seasoned software developer with over 5 years of experience, specializing in front-end development with React, team leadership, and product management. A problem solver passionate about delivering high-quality software, driving team performance, and constantly improving skills while staying up-to-date with the latest technologies and techniques. Dedicated to helping companies achieve their goals and delivering value through software by seeking new and more efficient ways to do so.