Geocoding
111 USD I'm looking for people who understand geoworkers and TP...
The key problem. Objects are flying to other labels and TP. The geocoding pyramid is not working.
Vibe coding is not helping... Friends vibe coders, this job is definitely not for you.
The task is to create a universal solution or fix the current one so that the address, district, and city are always displayed correctly on the map, or as close to that as possible (minimum error), so that it does not fly to another district or out of town. If there are those who really understand this, let me know.
There is a conditional object card that is located at a point on the map (geocoding), 85~% of the objects are correctly placed, while others are misplaced. Two assistants are involved: osm nomismatism and mapbox, but this does not help much. Districts are not drawn. An alternative is to draw polygons manually, but this will take a lot of time for each city. Therefore, I'm looking for something universal, who can advise?
-
4 days111 USD
4 days111 USD
Hello. I am working on geocoding/reverse geocoding, OSM (Nominatim), and Mapbox, as well as in a normal "admin panel" through polygons (PostGIS / point-in-polygon), which is why "shifts" to another area/out of town stop occurring.
According to your description, the problem is typical: you are trying to get the district/city from the geocoder's response (or reverse) — and this is heuristic, which is why 10-15% of points will always be incorrect, especially at borders and with incomplete addresses. The universal solution is to separate the tasks:
address → coordinates (forward geocoding),
coordinates → city/district (strictly by polygons, point-in-polygon), not "as returned by the geocoder." To assess the volume and propose an exact solution, I need:
10-20 examples of "departed" objects: original address/city + coordinates where it currently places them;
…
which cities/countries, and what data is guaranteed to be available for the card (city? district? postal code?);
where you store the data and what you use to draw the map (PostGIS available/not available, front-end stack).
After this, I will provide a work plan by stages and make a fix so that the address/city/district is determined as stably and reproducibly as possible, without manual drawing for each city.
-
3 days111 USD
1909 2 0 3 days111 USD
Hello!
I have thoroughly reviewed your task. The problem is familiar — I have encountered something similar when working with OSM Nominatim and Mapbox. Here is my vision:
Why objects "fly" to foreign markers:
Nominatim returns the centroid of the nearest found object, not the exact address. If the address is not found at the house level — it "jumps" to the center of the street, district, or even city. This is your 15% errors.
The problem with the geocoding pyramid — the classic fallback house → street → district → city requires validation at each level. It is necessary to check the importance and class/type in the Nominatim response, rather than just taking the first result.
…
Districts without polygons — I agree, drawing manually for each city is unrealistic. Solution: use administrative boundaries from OSM (relation boundaries, admin_level=9-10) or GeoJSON districts from open sources (data.gov.ua has boundaries for most cities). This will provide point-in-polygon validation without manual work.
My plan:
Audit the current pipeline: I will look at how exactly the fallback works and where accuracy is lost.
I will add validation by bounding box and type from Nominatim — filtering out "broad" results.
I will connect reverse geocoding for cross-checking (coordinates → address → comparison with the original).
For districts — I will pull OSM boundaries through Overpass API or ready GeoJSON, and perform point-in-polygon.
I will set up a confidence score: each result will receive an accuracy rating, so you can see where there is a 100% match and where manual verification is needed.
Result: 95-98% accuracy instead of the current 85%.
I have experience with Nominatim API, Mapbox Geocoding, Overpass API, Turf.js (geospatial operations), PostGIS.
Deadline: 2-3 days
Price: 5000 UAH
I am ready to discuss the details and take a look at the current code.
-
3 days111 USD
726 9 1 3 days111 USD
Hello! Your project looks very interesting. I am ready to start working immediately and ensure high quality.
-
2 days111 USD
1945 7 0 2 days111 USD
This is a typical story where the geocoder works "by line" without strict limitations and without checking against polygons. To estimate the volume and timelines, details are needed: Geography and which specific "districts" (level). Input data format: separate fields or one line. Examples of 5–10 problematic addresses + what Nominatim/Mapbox currently returns. Expected output: point, address, district, city, additional fields. Limitations on stack and infrastructure. If you want, I can audit the current pipeline and suggest a fix with tests on your set of cases.
Current freelance projects in the category AI & Machine Learning
AI agent for searching and analyzing a dataset of documents in the decisions register1. Context and Problem Target user: a specialist working with a large array of text documents and making decisions based on precedents. Essence of the problem: working with an open document registry takes excessively much time: search requires manual selection of keywords and… AI & Machine Learning ∙ 1 hour 23 minutes back ∙ 13 proposals |
Build a customer classification model1. There is client data in Mongo/SQL (approximately 20,000 entries with raw data). 2. It is necessary to build features and a classification model of clients into behavioral groups based on this data. 3. The project should be completed in Python. AI & Machine Learning, Python ∙ 1 day 15 hours back ∙ 30 proposals |
Integration of dental scanner modules into CRM
601 USD
We have developed a CRM system for interaction with dentists and laboratories. It is necessary to integrate services like iTero, Sirona, Medit, and others so that files are pulled automatically. AI & Machine Learning, Java ∙ 1 day 16 hours back ∙ 27 proposals |
Create a team of AI agentsI want to create a team of AI agents that will help in everyday life, control business processes, analyze reports, etc. AI & Machine Learning ∙ 1 day 19 hours back ∙ 31 proposals |
IT Automation of VAT Reporting
223 USD
It is necessary to develop a system for automating the transfer of sales data from the CRM to the accounting system Wafeq. The system should import bank and payment reports, automatically reconcile payments with invoices, generate invoices for VAT reporting, and minimize manual… AI & Machine Learning, Python ∙ 1 day 20 hours back ∙ 41 proposals |
Kyiv