AirStats Backend + Frontend
Budget: $750 – $1,500 USD
We're building the frontend right now and need an engineer that can build the backend in parallel.
It's a simple frontend (only 3 pages: signup/signin, form/popup, main dashboard).
The backend is also pretty simple - as long as you have experience building AI chatbots.
See the flow here:
https://www.figma.com/file/FcMAPzQYC1jc3qiqvYeF3v/Sherlock?type=whiteboard&node-id=0%3A1&t=kOfIKk7t4V2x2WO9-1
MORE IN-DEPTH BREAKDOWN
There's two main Backend Components:
-- 1 -- ACTION STEPS GENERATOR
We'll pass business metrics to a GPT-4 prompt and generate an array of priorities. Each item in the array will then get passed into a getActions() function that returns Action Steps for each item. The priorities + action steps will then get auto-populated as the first message from the chabot upon login (check the attachment to see what the Main Dashboard will look like)
From there, the conversation can be continued with the chabot - the chatbot should remember it's first message however.
-- 2 -- CHATBOT LOGIC
The chatbot's response logic is made up of:
(a) a GPT-4 prompt that labels the user's question
(b) a searchVectorIndex() function that takes the user's question and the GPT-4 label as arguments (the label informs the function of which index it should search for nearest neighbors).
(c) the NN matches are injected into a GPT-3.5 prompt to return an answer to the user's question
It's a simple frontend (only 3 pages: signup/signin, form/popup, main dashboard).
The backend is also pretty simple - as long as you have experience building AI chatbots.
See the flow here:
https://www.figma.com/file/FcMAPzQYC1jc3qiqvYeF3v/Sherlock?type=whiteboard&node-id=0%3A1&t=kOfIKk7t4V2x2WO9-1
MORE IN-DEPTH BREAKDOWN
There's two main Backend Components:
-- 1 -- ACTION STEPS GENERATOR
We'll pass business metrics to a GPT-4 prompt and generate an array of priorities. Each item in the array will then get passed into a getActions() function that returns Action Steps for each item. The priorities + action steps will then get auto-populated as the first message from the chabot upon login (check the attachment to see what the Main Dashboard will look like)
From there, the conversation can be continued with the chabot - the chatbot should remember it's first message however.
-- 2 -- CHATBOT LOGIC
The chatbot's response logic is made up of:
(a) a GPT-4 prompt that labels the user's question
(b) a searchVectorIndex() function that takes the user's question and the GPT-4 label as arguments (the label informs the function of which index it should search for nearest neighbors).
(c) the NN matches are injected into a GPT-3.5 prompt to return an answer to the user's question