ATLAS SEARCH/LUCENE IMPLEMENTATION WITH TYPESCRIPT

We are looking for an engineer experienced with Atlas Search / the underlying Apache Lucene.
Being comfortable working with Typescript and Atlas Search / MongoDB is a must.
*The task:* We provide a storefront containing 100k+ products from different companies.
These products have many properties.
name, description, brand, color, size, etc.
All of which ponder differently in our search results.
Our current issue is that for some queries, our atlas search implementation is suboptimal.
we suspect our set of rules could be built better, and we should also be able to leverage a different tokenizer that would work better for our use-cases.
We are looking for an experienced consultant to refactor our algorithm to fix some longstanding issues.
*What you will be given:* A document containing several queries that fail at finding the accurate results, along with the suspected reasons for failure Access to a github repo containing a small sample project with our search algorithm.
It takes an [login to view URL] file, reads it line by line, and then outputs csv files for each line in the input containing the search results for that line.
It can be used to understand / work on the algorithm, as well as check the results of the modifications.
If needed, several other documents containing additional failed queries, which might help understand the issues we face Access to the MongoDB database containing the products *What we expect:* A refactoring of the search algorithm (currently 400 lines), following atlas search best practices Improvements on the failed queries, wherever a fix is possible If a fix is not possible, detailed explanations of the steps we should take in order to achieve the desired results (add more data to the products, etc.) I have one screening question I require you to answer, either in detail or high level, but provide a proper explanation.
**Screening question:** In our setup we give an input query of "1.
mL".
We get the following wrong results.
*1st result:* Product Name.
Temperature Cable 400 Series Single, 0.
mL/1.
Foot For use with Temperature Probe SKU.
Reason#1 for Wrong Result.
.
Tokenization of 0.
mL leads to ‘match’ of “.
mL”, when it should not be split, and not returned as ‘exact match’ Reason#2 for Wrong Result.
.
Not ‘splitting’ input of “1.
mL” to find matches for either A.
“1.
mL“ and B.
“1.
mL” Reason#3 for ‘Wrong Result/a’.
should only record ‘exact matches’ to mL and 1.
ideally *2nd result:* Product Name.
Bipolar Forceps V.
Mueller® Malis 7 Inch Length Straight 1.
mm wide SKU.
Reason#1 for Wrong Result.
.
Incorrectly prioritizing match of 1.
mm OVER 1.
mL, when user searches “1.
mL” Reason#2 for Wrong Result.
.
Incorrectly fails to identify “1.
mL” as a perfect match! *Non-Specific Fails:* Fails to match any product that is “1.
mL” when user search 1.
mL Incorrectly matches only 41 results when user inputs “1.
mL”, but results when user inputs “1.
mL” *How it should work:* Ranking should go as follows.
- Top.
Exact matches for “1.
mL“ | Exact matches for “1.
mL” (equally) - Second.
Exact matches for exactly “1.
” | matches to exactly “mL” - Should not show.
Fuzzy matches for “1.
” or “mL” E.
.
should not show “125 ml” nor “0.
mL” - Should show similar results for two unique user inputs including “1.
mL” and “1.
mL” - Quantity or count of results - Similar rankings! I do not require a full solution, but respond to me in a way that shows your experience so I know it's worth having an interview.
Also include your rate in the answer.
Looking forward to meet a strong candidate! Apache Solr Typescript MongoDB Marketing do Motor de Busca JavaScript ID do Projeto.
# Sobre o projeto 50 propostas Aberto para ofertas Projeto remoto Ativo em 1 hora atrás