Data Querying Expert for Search Agent
Budget: €250 – €750 EUR
Looking for someone with hands-on experience in RAG and LLM-based data querying
We are building an LLM-based search agent to find suitable speakers for an event from a database of 700 speakers. Each speaker has a rate (costs to book this speaker), a ranking (how popular this speaker is), a category, a sub-category, keywords, a vita (html text about the speaker's life) and references (html text with previous bookings and feedback from other clients who booked this speaker).
Rate and ranking are integer values, category, sub-category and keywords are categorical while vita and references are unstructured text fields.
The bot should be able to combine classic filtering querys (like "give me all speakers with a budget between 10.000 and 12.000€", "all speakers below 5000€" etc) with partly semantic querys ("speakers whose category would match xyz") and fully semantic querys ("speakers who won a world cup", "speakers who worked in finance industry before" etc., this is information that is usually only given in the vita or references text field).
A possible example could be the following query:
"Give me all speakers with a budget below 6000€ that can hold a talk about motivation and leadership, that worked in a big tech company"
(The budget part is classic filtering, while the remaining part is partly- / fully semantic search.
Results should be ordered by rank, returning the speaker with the highest rank first.
The bot should also be able to briefly summarize the unstructured fields for the returned speakers and answer questions based on information given in the unstructured fields.
We would like to test the bot in a local environment (like just a jupyter notebook or a streamlit application), but it should be possible to transfer the bot to a web-based interface.
Possible technologies/framework could be:
- OpenAI API
- Google Cloud Platform / BigQuery / Vertex AI
- ElasticSearch
- langchain
If you know another tech stack that can solve this problem you can use it as well, of course!
When you apply to this job, briefly outline what tech stack you would use and why and give us a short summary of existing projects you worked on that might be relevant or similar to this project
We are building an LLM-based search agent to find suitable speakers for an event from a database of 700 speakers. Each speaker has a rate (costs to book this speaker), a ranking (how popular this speaker is), a category, a sub-category, keywords, a vita (html text about the speaker's life) and references (html text with previous bookings and feedback from other clients who booked this speaker).
Rate and ranking are integer values, category, sub-category and keywords are categorical while vita and references are unstructured text fields.
The bot should be able to combine classic filtering querys (like "give me all speakers with a budget between 10.000 and 12.000€", "all speakers below 5000€" etc) with partly semantic querys ("speakers whose category would match xyz") and fully semantic querys ("speakers who won a world cup", "speakers who worked in finance industry before" etc., this is information that is usually only given in the vita or references text field).
A possible example could be the following query:
"Give me all speakers with a budget below 6000€ that can hold a talk about motivation and leadership, that worked in a big tech company"
(The budget part is classic filtering, while the remaining part is partly- / fully semantic search.
Results should be ordered by rank, returning the speaker with the highest rank first.
The bot should also be able to briefly summarize the unstructured fields for the returned speakers and answer questions based on information given in the unstructured fields.
We would like to test the bot in a local environment (like just a jupyter notebook or a streamlit application), but it should be possible to transfer the bot to a web-based interface.
Possible technologies/framework could be:
- OpenAI API
- Google Cloud Platform / BigQuery / Vertex AI
- ElasticSearch
- langchain
If you know another tech stack that can solve this problem you can use it as well, of course!
When you apply to this job, briefly outline what tech stack you would use and why and give us a short summary of existing projects you worked on that might be relevant or similar to this project