🏫
Obvious University
Website
  • 👋Welcome to Obvious University!
  • Strategy
    • Sprints
      • 1️⃣Map
      • 2️⃣Sketch
      • 3️⃣Decide
      • 4️⃣Prototype
      • 5️⃣Test
    • Benchmarking
    • Research
      • 1️⃣Research guide
      • 2️⃣How to recruit users
      • 3️⃣How to conduct an interview well
      • 4️⃣How to take notes
      • 5️⃣How to prep for remote research
      • 6️⃣How to throw a watch party
      • 7️⃣How to create artefacts
  • Working with Features
    • Building with AI
      • 1️⃣Understand the tech
      • 2️⃣Map your product
      • 3️⃣Build a proof of concept
      • 4️⃣LLM Inputs
      • 5️⃣LLM Responses
    • Building Help and Support
      • 1️⃣How to scope a support experience
      • 2️⃣How to design discovery for support
      • 3️⃣How to design a support centre
      • 4️⃣How to write good support articles
  • Product Design
    • Microcopy
      • 1️⃣How to write well
      • 2️⃣How to write phrases
      • 3️⃣How to write messages
      • 4️⃣How to create a voice
    • Typography
      • 1️⃣How to compose type
      • 2️⃣How to create a type scale
      • 3️⃣How to pick typefaces
      • 4️⃣How to pair typefaces
    • Design System
      • 1️⃣Introduction to design systems
      • 2️⃣How to audit a design system
      • 3️⃣How to run a design system pilot
      • 4️⃣How to set up a design foundation
      • 5️⃣How to build components
      • 6️⃣How to document a design system
      • 7️⃣How to enable adoption and govern a design system
    • Mobile Engineering
      • 1️⃣Trunk based development
      • 2️⃣Agile development terminology
      • 3️⃣Git commit messages
      • 4️⃣Code review and pull requests
      • 5️⃣Readings
  • Delivery
    • Project Management
    • Collaboration
  • Hiring and Growth
    • Growth
      • 1️⃣Design growth framework
      • 2️⃣How to give ongoing feedback
      • 3️⃣How to check-in every quarter
      • 4️⃣How to address underperformance
      • 5️⃣FAQs
    • Hiring and careers
      • 1️⃣The Hiring Process
      • 2️⃣Diverse and Inclusive Hiring
  • PEOPLE EXPERIENCE
    • Benefits and Perks
      • 1️⃣Paid time off
      • 2️⃣Insurance and healthcare
      • 3️⃣Continuing education
      • 4️⃣Speaking at conferences
    • Starting at Obvious
      • 1️⃣Introducing Obvious
      • 2️⃣Set up your workspace
      • 3️⃣Onboarding
      • 4️⃣Finances
      • 5️⃣Code of Conduct
    • Employment policies
      • 1️⃣Equal opportunity employment
      • 2️⃣At-will employment
      • 3️⃣Employee records and privacy
      • 4️⃣Prevention of sexual harassment
      • 5️⃣Drugs and alcohol
      • 6️⃣Fraternisation
      • 7️⃣Non-compete and non-solicitation
      • 8️⃣Non-disclosure
Powered by GitBook
On this page
  • 👋 Introduction
  • 1️⃣ Inputs
  • 2️⃣ Outputs
  • Using an LLM
  • Iterate on prompts
  • Fine-tune your model
  1. Working with Features
  2. Building with AI

Build a proof of concept

PreviousMap your productNextLLM Inputs

Last updated 1 year ago

Assumed audience: You’re a product designer working on an LLM powered feature. You’re familiar with how LLMs work. Use this as a starting point for further exploration and innovation.

👋 Introduction

You have a solution in mind, it’s time to start testing feasibility and support your engineering team. Here’s how you can bring in the required product context into the building process.

1️⃣ Inputs

Identify the sources for the LLM to pull from.

Some questions to get you started—

  1. What type of queries would you anticipate from users?

  2. Do you have a knowledge base that will act as a source? Any other form of documentation?

  3. Does it need additional interaction level information about the user’s behaviour?

  4. Is there going to be a collaborative back-and-forth between the user and the LLM? If so, how should you retain context in an LLM<>user collaboration session?

2️⃣ Outputs

Articulate what a meaningful response looks like for your use case. Identify what kind of fine-tuning or prompting you’d need to get to this response reliably.

Using an LLM

Bring in stakeholders who have context of the product to collaborate on the prompts and examples.

Iterate on prompts

Fine-tune your model

Do this if a prompt is not effective. It will improve accuracy, give a deeper understanding of the product space, and be able to respond to a wide input variety.

📣 “Fine-tuning improves on few-shot learning by training on many more examples than can fit in the prompt, letting you achieve better results on a wide number of tasks. Once a model has been fine-tuned, you won't need to provide as many examples in the prompt. This saves costs and enables lower-latency requests.”

Create a dataset of ideal responses. Creating good training data is a design and copy responsibility. Generating high quality example data is critical to build the best experiences. Learn more about crafting responses

Keep them specific and context rich. Prompt engineering uses natural language so it’s quite accessible, and is core to the product experience. If you’re using OpenAI, see how far you can get with just testing out your idea on the before you actually start pulling from all your sources.

From

3️⃣
here.
Playground
Fine-tuning the OpenAI API