API Technical Specs
1. Plan your workflow
It is important to understand how each part of the system works together so you can optimize your integration. If you are considering building your own UI please refer to the User Interface section for an in-depth look at what you will be undertaking. Remember, the decision makers that chose Sovren's AI Matching did so after seeing our UI during the demo. Simply showing a results page and populating it with static data from the API call is not a proper implementation. Please reach out to email@example.com to discuss this in more depth and get recommendations tailored to your use case.
2. Create your indexes
There are a few different ways to design your indexing strategy; for recommendations, refer to Indexing Strategy.
3. What data to index?
While you will likely index resumes and jobs, there are more data points to consider. Sovren's AI Matching Engine allows adding tags to each document that can later be used for searching and filtering. Typically, customers use this to add information not found on a resume such as status, willingness to travel, availability date, etc. The complete documentation for Custom Values can be found here.
4. Backfilling your indexes
Typically, customers have a backlog of documents to add to the index. This can be done using the Batch Utility found on our GitHub. Only index documents that will be relevant for searching/matching. For example, documents that are more than 18 months old, or that are inactive in your system, should not be added to the AI Matching indexes. Reach out to firstname.lastname@example.org if you have questions around what data you should index.
5. Category Weights
The largest value add that Sovren's AI Matching provides is the ability to return the best matches with a simple API call. Integrators do not have to build complex faceted search screens and train recruiters how to build complex queries, but rather leverage the intelligence of Sovren's Matching to do the heavy lifting.
One of the biggest differentiators between our engine and others is that we allow you to disagree with the results. Our AI Matching evaluates each document across 8 categories and then uses a weighting system (category weights) to calculate the rolled-up scores. Integrators have complete control over the weights and can change them for each transaction. This allows you to tailor each match to the candidate, job, or client you're trying to match for. For example, if you were trying to fill two positions, one for a data scientist and another for a forklift driver, you would not look for the same criteria. Our engine works the same. Sovren will calculate suggested weights based on the data input but allow the integrator to overwrite those for superior control over the results.
Sovren uses very transparent scoring to rank and sort the match results. There are a few different scores that are returned, so it is important to understand what each one represents and what you should be using. For more details on scoring refer to the Scores documentation.