1
1 Comment

Extracting problems form job posts. Wanna buy it?

TL'DR
I'm selling access to my un-deployed, not finished, proof of concept AI tool that analyses job posts from a popular freelancing platform to try and identify the problem the client is trying to solve.

How can you buy it?

  1. Leave a comment with your twitter handle/email and I'll DM you.
  2. I don't have a price for it yet. We'll negotiate, but probably $50+.

STATS
Ran and refined it for about 1 week.

Processed jobs: ~2k
AI Identified problems: ~200
Problems that I was able to identify: 19

Disclaimer:

  1. As i said this is not deployed. The LLAMA model is running on my macbook so throughput is not high.

19 Problems I found
πŸ€– AI: The job description is looking for an expert in Notion and HubSpot to help craft an integrated Notion Workspace view into their Customer Success pipeline. The integration would take data from HubSpot, their own API, and create an automatically populated view in Notion indicating customer health. The problem is to find a freelancer or consultant with previous experience in both Notion and HubSpot who can help with this integration.
πŸ‘¨ HUMAN: It's not easy to to add in data from new sources to a notion view and have it automated.


πŸ€– AI: The job description is asking for a service that turns large documents (stories or movie scripts) into a series of nodes which can be altered and introspected using the repository linked above or something very similar. The service should generate sequential timelines for each character and their physical state, actions, dialogue, and environment, and allow for editing and exporting of JSON files. The problem is to find a solution that can efficiently and accurately convert large documents into a graph visualization with the graph library API, and provide a Python API for uploading or exporting.
πŸ‘¨ HUMAN: A tool is needed to extract data from unstructured docs and graph them.


πŸ€– AI: The problem is to develop a document classification and extraction system using AWS tools such as SageMaker, Comprehend, and Textract to streamline the document management system. The system should automatically classify new documents based on predetermined criteria and extract relevant texts to be inserted into a database. The project requires proficiency in AWS services, machine learning, natural language processing, database management, and integration skills. The budget for the project is $4,000.
πŸ‘¨ HUMAN: It's hard to structure a large set of documents stored different non-digital formats in order to organise data.


πŸ€– AI: The problem is finding the best online resources that teach how to use basic AI tools for productivity at work.
πŸ‘¨ HUMAN: Office worms have no idea how to use AI to make their jobs eaiser. Employers want to teach them to use basic AI tools improve productivity.


πŸ€– AI: The problem is to automate the human evaluation and feedback process for a case study exam preparation company. The company is looking for an AI solution that can analyze and understand the background document, match it to the client's written submission, and provide detailed and constructive feedback and scoring based on the company's scoring criteria/grid.
πŸ‘¨ HUMAN: Test evaluators need a lot of time to evaluate multiple tests based on case studies. The evaluation could be sped up with ChatGPT assistant


πŸ€– AI:
πŸ‘¨ HUMAN: Using Vertex AI businesses want to ask their collected and structured data real questions for insights.


πŸ€– AI:
πŸ‘¨ HUMAN: Professional photoshoots videos or photos sometimes include logos of brands. It takes a lot of time to identify the logos and blur/hide them out in a way that it does not compromise the quality of the photo/video.


πŸ€– AI:
πŸ‘¨ HUMAN: UK businesses that want to enter the US market are having trouble signing up for FEIN.


πŸ€– AI: Pre-filtered as a service job keywords
πŸ‘¨ HUMAN: Seems that Klaviyo does not produce insights that are advanced enough for certain users. But they can download raw performance data.


πŸ€– AI:
πŸ‘¨ HUMAN: There is no AI model that is roboust enough to structure unstructured data into system wide structures. In this example: "We retrieve information from a variety of souces, some of which is quite messy and difficult to standardize. We are looking to fine-tune an Azure OpenAI model on a set of pre-formatted documents so that we can feed in future documents and have it be processed in the same way, thus creating a standardized template for our documents. We are looking for an enginner with experience working Azure OpenAi and other LLMs along with experience fine-tuneing and/or working embeddings."


πŸ€– AI:
πŸ‘¨ HUMAN: Early founders don't know how to create an audience. I can relate to that too.


πŸ€– AI:
πŸ‘¨ HUMAN: When you want to invest in real estate abroad it's really hard to understand what your costs and how the regulations work. Here by invest I mean buying an apartment/small home.


πŸ€– AI:
πŸ‘¨ HUMAN: Potentially lexoffice does not have automation integrations for auto invoicing and maybe there is a way to integrate it with more tools that are popular among lex office users.


πŸ€– AI:
πŸ‘¨ HUMAN: Business owner uses his own internal table to track all the financial information. Now he wants a deep understanding on the financial performance of his company. He cannot leverage any saas products that could give him that so he needs to hire a financial analyst.


πŸ€– AI:
πŸ‘¨ HUMAN: US expats in germany that are doing freelancing have trouble with US tax accounting.


πŸ€– AI:
πŸ‘¨ HUMAN: Controlling telescopes/telescope mounts to track moving objects on the sky is not accessible for hobbyists.


πŸ€– AI:
πŸ‘¨ HUMAN: I don't know enough about the 4g gateways and mikrotik to understand what the client is trying to solve. But there might be smth there.


πŸ€– AI:
πŸ‘¨ HUMAN: Old established businesses are having trouble refurbishing old handbooks/documentation/policy documentation.


πŸ€– AI: Pre-filtered as a service job keywords
πŸ‘¨ HUMAN: It is hard to find backlinking opportunities that are of good quality and win/win for both sides.


on September 21, 2023