Job Twin brief — UpSkillZone AI
Job Twin 1 — Inference fundamentals
Scenario
Scenario
You're shipping a streaming chat endpoint. Build it with proper backpressure, partial-output handling, and a graceful timeout.
Time-box: 6 hours. Submit a runnable repository plus a tradeoff writeup.
Deliverables
Deliverables
- Streaming endpoint — token-by-token SSE or chunked transfer.
- Backpressure path — slow-consumer test that doesn't OOM the server.
- Tradeoff writeup — 300–600 words on cancellation, retries, and timeouts.
Materials
Time-box
6 hours
Server-authoritative clock. The deadline is hard; auto-save does not extend it.
Submission modes
- repo_url
- code_editor
The first mode listed is the default on the submit screen.
Rubric
Each dimension scored on [0.0, 1.0] in 0.05 increments. The overall score is the weighted average; pass at 70%.
| Dimension | Weight | What it tests |
|---|---|---|
Problem decomposition problem_decomposition | 15% | Breaks streaming concerns into testable layers. |
Streaming correctness streaming_correctness | 20% | Tokens flush incrementally; no full-response buffering. |
Backpressure handling backpressure_handling | 15% | Slow consumers do not exhaust memory or block peers. |
Error handling error_handling | 15% | Timeouts, cancellations, and upstream failures are graceful. |
Code quality code_quality | 15% | Readable, deterministic, runs. |
Docs quality docs_quality | 10% | README explains how to run + extend. |
Tradeoff writeup tradeoff_writeup | 10% | Defends choices on cancellation, retries, and timeouts. |
Failure modes
Self-checks the learner answers before submit. Critical checks block submission unless explicitly forced; the force flag is then surfaced to the mentor.
- F1
Does the endpoint stream tokens, not buffer the full response?
critical
- F2
Does a slow-consumer test pass without OOMing the server?
reflective
- F3
Are upstream timeouts surfaced as a clean error frame?
reflective
Skill assertions on offer
On a passing review the mentor selects a subset of these to assert, with an asserted weight bounded by the per-skill ceiling shown below.
- max weight 1.00
llm.inference.streaming-design
LLM inference — streaming design
- max weight 0.80
llm.api.error-handling
LLM API — error handling
Mentor SLA
72h
From mentor claim to signoff.
Pass threshold
70%
Weighted-average overall score.
Re-attempts
1
Higher of the two scores flows to the credential.
Start this twin
The clock starts when you press start. Read the brief above first. You will be asked to sign in if you have not already.