Job search
Uncountable Take-Home Assignment: Plan the 90-Minute Build and the Engineering Presentation
Separate effort time, delivery deadline, runnable code, presentation evidence and the later Python SDK debugging stage.
Uncountable’s full-time take-home is a short build followed by an engineering conversation. Its public guide asks for around 90 minutes of work, with time tracked, then code and run instructions in advance of a presentation. Plan one useful feature slice, a runnable handoff and an honest account of what remains.
This guide reviews current official public instructions checked October 23, 2026. We did not receive a candidate assignment, build an employer’s exercise or attend an interview. The two completed planning examples below are original fiction. Keep your actual invitation alongside them; for other applications, browse recent roles on LandOffer.ai.

Read the 90-minute request in its actual context
The Uncountable Engineering Interview Guide places the take-home under Full-Time Only. It says the interviewer emails the description and materials, allows work at your own pace and asks you to track roughly 90 minutes. The public wording establishes an approximate effort expectation; it does not describe a browser countdown or automatic submission cutoff.
The same guide asks candidates to send code and instructions for running it beforehand, then present to engineers during or before the onsite. The published onsite is 120 minutes and includes past work, technical discussion and live Python SDK debugging in an online IDE. A scheduling agenda supplies further details. Do not assume those 120 minutes are all available for your take-home presentation.
Before building, extract the requested feature, supplied data, permitted tools, expected delivery channel and submission deadline from the genuine instructions. Record effort time and the calendar delivery deadline separately. A 90-minute build budget does not tell you whether the email is due tonight or two days before the panel.
If the invitation and public guide differ, resolve that discrepancy with the coordinator through the existing invitation. Do not silently choose whichever version gives you more time. Similarly, the public page does not establish a universal repository host, ZIP format, deployment provider or minimum number of tests. Those are questions for the actual assignment, not gaps to fill with assumptions.
Choose a feature slice before choosing extra infrastructure
Start with the smallest behavior that lets someone complete the core task. Write one acceptance sentence in ordinary language: who acts, what changes and what visible result establishes success. Then identify one failure state that would otherwise confuse the user. Use that sentence to decide what the demonstration must show.
Separate required scope from optional improvements. Authentication, production hosting, elaborate styling and a new database can consume the entire budget without making the requested behavior easier to evaluate. Include them when the instructions require them or when the core feature depends on them; otherwise record the tradeoff rather than quietly expanding the project.
Reserve time for a clean run and the handoff. Working code that only starts in your existing terminal session is harder to discuss than a smaller implementation with reproducible steps. Note the actual runtime, installation command, start command and any local fixture assumptions. Never include secrets or private employer data in the package.
Testing should support the slice. A normal path, a meaningful invalid input and an important boundary may reveal more than many superficial checks. Choose tests based on the behavior you implemented. Passing a handful of local checks would not prove production readiness, security or compatibility with every reviewer’s machine.
Completed build-budget record: a bounded equipment planner
Fictional example one: Priya invents a practice project called BenchSlot, a local interface for reserving shared workshop equipment. It is not an Uncountable prompt or a solution to its private exercise. Her acceptance sentence is: “A workshop member can request an available time slot and receive a clear conflict message if that slot is already reserved.”
Her completed plan deliberately uses a local fixture and one browser view. It excludes accounts, payments, recurring reservations and cloud deployment. These exclusions belong to this invented practice scope; they are not permissions to omit requirements in a real assignment.
| Effort interval | Minutes | Completed planning allocation and evidence target |
|---|---|---|
| 0–8 | 8 | Read the invented brief; record slot granularity and the meaning of a conflict |
| 8–18 | 10 | Sketch available, selected and conflicting states; choose a local fixture |
| 18–53 | 35 | Implement selection, reservation and conflict feedback as one vertical slice |
| 53–68 | 15 | Check a free slot, a duplicate request and an adjacent non-overlapping slot |
| 68–80 | 12 | Write setup/run instructions and try a clean local start |
| 80–90 | 10 | Capture limitations, actual time and a short presentation outline |
| Total | 90 | Planning arithmetic completed; no software execution asserted |
Her decision record names the main ambiguity: whether adjacent reservations conflict. For this practice plan she stipulates half-open intervals, so a booking ending at 10:00 does not block one starting at 10:00. In an actual take-home, that assumption would be documented or clarified rather than presented as an employer requirement.
Priya also completes her proposed handoff inventory: source directory, dependency manifest, README, local fixture and a brief limitations note. The README must state the exact commands after implementation establishes them; this plan does not invent commands for code that does not exist. “Clean start checked” remains unconfirmed because this example is a planning artifact.
Her final scope decision is specific: keep conflict handling; defer recurring reservations. The reason is that recurrence adds date rules and editing behavior beyond the acceptance sentence. Her allocation reserves 22 minutes for the clean-start instructions, limitations and presentation outline. It does not claim that the imagined tests passed or that any code was submitted.
Make the code handoff usable without inventing a delivery rule
Package the implementation according to the actual instructions. A reviewer should be able to locate the entry point, understand the required environment and identify the data needed for a first run. If you rely on environment variables, list names and safe setup guidance without transmitting credentials.
Explain what a successful run should show. “The app starts” is less useful than a short path through the implemented behavior and one known limitation. Keep the description aligned with the actual version you send. A screenshot can illustrate the interface, but it cannot replace runnable source when code is requested.
Check whether a dependency on a paid service, external API or account is necessary. If the instructions do not settle that dependency, clarify before making the reviewer obtain access. A local substitute can be appropriate only when it preserves the required behavior and its limitations are disclosed. Do not claim a mocked response proves a real integration works.
Record delivery and receipt separately. Sending a link does not establish that permissions allow access, and an accessible repository does not confirm that a coordinator has received it. Use the required channel and retain the actual delivery evidence. This guide does not establish a default deadline or authorize contacting anyone on your behalf.
Prepare an engineering presentation around decisions and evidence
Lead with the user problem and the feature that works. Show the shortest demonstration that supports your acceptance sentence. Then explain the implementation choice that mattered most: a data boundary, a state transition, an error path or a deliberately deferred requirement. Let the audience see why you spent the time where you did.
Distinguish implemented behavior, checks actually run and proposed future work. A prototype may be useful while lacking persistence, concurrency handling or accessibility review. Name the relevant limitation without burying the working feature under a list of everything a production system could eventually need.
If the panel asks about an unfinished part, describe its current state and the next concrete step. Do not quietly repair it during a claimed final demonstration and imply it was already complete. Keep a small local fixture or other permitted fallback ready for presentation problems, while identifying it honestly as a fallback.

Independent presentation case: explain a deliberate deferral
Separate fictional example two: Mateo’s original practice project, SampleRoute, routes laboratory sample labels to a receiving desk. Its imagined implementation lets a user select a destination and see a readable confirmation. Unlike Priya’s reservation planner, this case addresses how to discuss a completed prototype’s tradeoffs. All behavior below is stipulated fiction, not observed execution or Uncountable assessment content.
Mateo completes this presentation record before an invented eight-minute rehearsal. Eight minutes is his personal rehearsal target, not an employer-assigned slot.
| Rehearsal segment | Minutes | Filled presentation content |
|---|---|---|
| Problem and user | 1 | Receiving staff need a clear destination confirmation before attaching a sample label |
| Core demonstration | 2 | Select the fictional microscopy desk, confirm the destination and show the resulting label preview |
| One engineering decision | 2 | Keep destination identifiers separate from display labels so a renamed desk does not require rewriting stored references |
| Evidence and limit | 2 | Scenario stipulates checks for a known destination and a missing selection; no printer, persistence or concurrent-user test is established |
| Next step and questions | 1 | Add a printer adapter behind a defined interface after confirming the required device and failure behavior |
| Total | 8 | Rehearsal plan completed; no actual panel or recorded rehearsal claimed |
His completed tradeoff statement reads: “The prototype makes destination selection and preview understandable. I deferred printer integration because the practice brief supplies no device contract. I would first define the input, successful acknowledgement and recoverable failure, then test against the specified adapter. The current preview does not prove that a physical label prints.”
His difficult-question note is also filled: “If two desks share a display name, I would preserve distinct identifiers and show enough location context for the user to choose. That behavior is proposed, not demonstrated here. I would confirm the intended naming rules before changing the interface.” This gives the conversation a concrete next decision without upgrading an idea into shipped behavior.
The completed artifact is a presentation outline with each claim tied to its stated evidence. Mateo has a reason for the deferral, a next investigation and a truthful account of the stipulated checks. He has not claimed production reliability, a successful hardware integration or approval from an actual interviewer.
Prepare separately for the Python SDK debugging stage
The public guide’s Python debugging stage deserves its own preparation. A general preference for another coding language does not remove that explicitly described activity. Review the agenda when it arrives and identify what environment or preparation the coordinator actually specifies.
The current Uncountable Python SDK overview describes a wrapper around its REST API, with client authentication, error propagation and pagination. Reading that public overview can help you recognize interface concepts. It does not establish the interview’s installed package version, supplied credentials, dataset or exact bug. Do not create a production integration merely to rehearse an interview.
Practice explaining how you narrow a failure: identify the failing line, inspect the relevant value and type, compare expected with actual behavior, make one change and rerun the smallest useful check. Separate an invalid local value from a remote response or an incomplete result set. Explain what evidence would distinguish the possibilities rather than guessing an unseen API contract.
Python’s official debugger documentation explains breakpoints, stepping and stack inspection. Those are useful general concepts; we did not verify that this particular online interview environment exposes pdb or permits installing anything. Prepare to explain your reasoning with the tools the interviewer actually provides.
Finally, keep tool permission stage-specific. A statement about optional coding tools in one interview section does not settle every take-home or debugging rule. Follow the genuine instructions and clarify missing permission before use. With your effort record, runnable handoff and presentation boundaries prepared, return to your broader search through recent LandOffer.ai roles.
Sources and Further Reading
Official public sources checked October 23, 2026. Employer guide and current SDK overview were read in the browser; no private assignment, account, API request, code submission or interview was accessed.
- Uncountable: Engineering Interview Guide, current public process; no visible revision date.
- Uncountable: Python SDK overview, current public documentation; not a verified interview runtime specification.
- Python: pdb debugger documentation, general debugging reference.


