Skip to main content

Job Benchmark

  • Last updated:

External Benchmark

We use the title and the description of the selected job offer along with the business type of the client to find information on the web about the skills needed for the business by doing simple web search like : {title}'s skills} or {business_type}'s trending skills and summarizing the results. We also search open positions related to the title and extract their skills.

With all this data, we then proceed to sort skills by trending and emerging ones with custom scores (taking into account how many skills we extracted, what we found in web search and LLM reasoning).

Using a LLM the offer description and title are translated in english for better usability in search and LLM. We ask for different versions of the titles to extand our searching range and replace the description by explicit tasks and responsibilites infer from the description.

We use a web search engine (like google, bing...) to perform a list of searches. We do :

  • "{business_type} skills"
  • "{business_type} trending skills"
  • "{market_offer.english_title} skills"
  • "{market_offer.english_title} trending skills"
  • "{market_offer.title}"

We select the top 3 articles for each search (except the last one where we select the top 5 articles). From the web search engine overview of the articles, we ask a LLM to do a concise summary of each articles.

We use a special search in a search engine specialized in finding open positions. We search the offer's original title and aim for 16 open positions. If we don't find 16, we search for the alternative title generated at the previous step, until we got 16 open positions.

Finally, from each summary we run a LLM prompt to extract skills. We ask in the meantime if the LLM thinks the skill is specific, mandatory, trending or in demand for this job. We provide all the previously scrapped articles as reference.

Then we count all skills extracted at the previous step and group them by similarity. We check sort them by their popularity and trending scores for the "Top" and "Emerging" sections respectively.

Skill Information

We use web searches and LLM reasoning to provide a skill description and reasons why this skill is useful for the offer in the client's industry.

We use a web search engine (like google, bing...) to search for :

  • "{skill_name} in {business_type} industry"

We select the top 4 articles. From the web search engine overview of the articles, we ask a LLM to produce a skill description and reasons why this skill is useful for the offer in the client's industry.

Skill Articles

We use web searches and LLM reasoning to provide a list of interesting articles related to the skill in the client's business.

We use a web search engine (like google, bing...) to search for :

  • "{skill_name} in {business_type} industry"

We select the top 8 articles. From the web search engine overview of the articles, we ask a LLM to produce tell if the article is interesting or not.

We take the three most relevant articles, search a random google image related to the skill name for each of them, and present them to the client.