At Zesty.ai, Back-end engineers:
- design high-performance, reliable, cloud-native distributed systems
- work with many kinds of data and imagery, and
- build and extend Zesty.ai’s platform, services, and internal tools.
This test serves as an opportunity to demonstrate comfort and skills with developing back-end services as well as addressing difficult unknown problems. The content of this test is designed to cover challenges that are similar to ones one would encounter working at zesty.ai. This test should be challenging, but also fun. It's meant to also help candidates evaluate if they would enjoy the kind of problems one might face working on our stack.
The goal is to create a RESTful API (see API Specification) written in Python that can locate and manipulate images and geographical data. You'll use property data stored in an SQL database and images stored in cloud storage (See Feature List). The API should be packaged as a containerized service. A test property database and images are provided for you (see Setup).
Note that some features are more difficult than others, and you will be evaluated on more than just the number of features completed. Quality is preferred over quantity. Design, organize, and comment your code as you would a typical production project. Be prepared to explain any decisions you made.
-
Display: API endpoint to display an image by property ID. Given a propertyId as input, find the image URL from the database, download it from Google Cloud Storage and return a JPEG image.
-
Find: API endpoint to search properties within a geographical area. Take a GeoJSON object geoJson and a search radius distance (in meters) as inputs. Return all property IDs that are within distance meters of geoJson. Use the
geocode_geofield of thepropertytable for the query. -
Display Plus: An alternative version of the first API endpoint (Display) to also overlay the geocode (as a marker - such as a small circle or triangle), parcel, and/or building (
geocode_geo,parcel_geo, andbuilding_geofields in the database, respectively) on the image. Optionally, add parameters for the color of each overlay, or use a default for each. -
Statistics: API endpoint to calculate geographic data about all properties within a given distance from a reference property. Take propertyId and distance (in meters) as inputs. The API should return the following:
- parcel area (meters squared)
- building area (meters squared)
- building distance to center (meters). Distance to center is the distance from the
building to the
geocode_geofield in the property table - zone density (percentage). Create a "zone" geography, which is a buffer of distance
meters around the
geocode_geopoint. Then, calculate the percentage of that zone geography which has buildings in it.
-
Freestyle: Based on the other features, you should have a feel for the kind
of features this API implements. If you have other cool ideas of things to add that aren't listed here, we'd love to see them.
Install docker to run the example database.
The deliverable should be in Python 3.x and be made available as a Docker image.
From the repo root folder, run docker compose up -d to start the PostgreSQL database needed
for this example. The database server will be exposed on port 5555. If this port is not available,
feel free to change the port in the docker-compose.yml file.
In the test database, there is a single table called properties (with 5 sample rows), where each row
represents a property or address. There are three geography fields and one field with an image URL
pointing to an image on Google Cloud Storage.
If you are not familiar with PostgreSQL or PostGIS, you may need to read up beforehand.
The API must adhere to the following API specification.
Note that this includes the specification for all endpoints. Ignore the ones for features that are not implemented in the submission.
Fetches and displays property image (as JPEG) by ID
example: GET localhost:1235/display/f853874999424ad2a5b6f37af6b56610
id| description: Property ID | type: string | required: true
JPEG image
Finds properties within X meters away from provided geojson point.
example: POST localhost:1235/find
JSON object with the following properties
location| geojson object representing point to search from | required: true | validation: geojson | type: objectdistance| distance, in meters, to search fromlocation| required: true | validation: greater than 0, less than some reasonable number | type: float
example:
{
"location": {
"type": "Point",
"coordinates": [-80.0782213, 26.8849731]
},
"distance": 500
}
JSON array with objects containing at least the following fields (you may include more if you think they would be helpful)
property_id| ID of property objectdistance_m| actual distance from inputdistance
Fetches and displays property image (as JPEG) by ID, with feature overlays drawn on image
example: GET localhost:1235/display/f853874999424ad2a5b6f37af6b56610/overlays?building=green&parcel=orange
id| description: Property ID | type: string | required: truegeocode_color| description: color to use for geocode overlay. empty to use default color (default should be different from other feature defaults) | type: string | required: false | validation: enum (ex. "red", "green", "orange") or hex string (ex: "FF0040")parcel_color| description: color to use for parcel overlay. empty to use default color (default should be different from other feature defaults) | type: string | required: false | validation: enum (ex. "red", "green", "orange") or hex string (ex: "FF0040")building_color| description: color building overlay. empty to use default color (default should be different from other feature defaults) | type: string | required: false | validation: enum() ex. "red", "green", "orange"
JPEG image
Returns various statistics for parcels and buildings found X meters around the requested property
example: GET localhost:1235/statistics/f853874999424ad2a5b6f37af6b56610?zone_size_m=10
id| description: Property ID | type: string | required: true | validation: length greater than 0zone_size_m| description: Buffer distance used for calculating zones | type: integer | required: true | validation: greater than 0
JSON object including following fields
parcel_area_sqm| description: total area of the property's parcel, in square meters | type: floatbuilding_area_sqm| description: total area of building, in square meters | type: floatbuilding_distance_m| description: distance from the centroid of the property, to the centroid of building, in meters | type: floatzone_density| description: density (%) of building's area to zone area (the zone is thegeocode_geowith a buffer/circle with the radius ofzone_size_minput) | type: float
Send us your completed application's code by email, or create and give us access to a new private GitHub repository.
Include instructions on how to run your app, and a list of what features you implemented. Add any comments or things you want the reviewer to consider when looking at your submission. You don't need to be too detailed, as there will likely be a review done with you where you can explain what you've done.
Your submission will be scored out of 150 using the following scorecard:
| Feature | Max score |
|---|---|
| Display | 10 |
| Find | 20 |
| Display Plus | 40 |
| Statistics | 40 |
| Freestyle | 20 |
| Containerization | 20 |
| Total | 150 |
Note that the points attributed to each feature are not all-or-nothing. Partial points will be attributed for incomplete features. If you can't complete a feature, you're still encouraged to submit code that shows signs of your progress. It's possible to get a passing grade without implementing all features, as quality is preferred over quantity. Extra points, not included in table above, will be given for going above expectations quality-wise .
There is intentionally no time limit set on this test. Spend as much time as you feel you need to create a deliverable you are proud of, but like in a real project make sure to balance value and effort.
Yes. Knowing when to ask for help is a skill that is equally important as being able to solve things yourself.
Sure. As long as the projects adequately demonstrate your technical prowess. We will evaluate such submissions and let you know if we still need anything further.
We think Zesty.ai has a lot to offer. We do exciting work and use cutting-edge technologies. We would be happy to schedule time to talk to one of our engineers to give you more details and let you ask questions before you start the test.