The setting

Bluebell Health runs a telehealth service where patients book video visits with clinicians across four states. You are a frontend engineer on the patient web app, a small team where the interface is the product.

Every mission on this path happens at the same company, so context carries over the way it does in a real job: the data you cleaned in mission two is the data the finance lead questions in mission four.

The missions

1. accessibilitystarter

Audit the Bluebell booking form for accessibility

A patient who uses the NVDA screen reader emailed Bluebell to say she could not finish booking a visit, and a state Medicaid contract renewal now requires WCAG 2.1 AA conformance evidence. The booking form is the first page under review.

You deliver: An accessibility audit of the booking form mapping each violation to a WCAG criterion and a concrete fix.

Scored on: Finds the violations, Contrast checked, Explains the patient's report, Concrete, correct fixes.

Working from: booking_form.md, patient_email.md.

2. responsive layoutstarter

Make the visit summary usable on a 375 pixel phone

Bluebell's visit summary page, the one patients open right after a video call, was designed on desktop and it shows: 62 percent of visits now end on a phone, and support keeps getting screenshots of a squashed, overlapping page from smaller iPhones and budget Android devices.

You deliver: A responsive layout plan that fixes each breaking CSS rule and maps every support ticket to its cause.

Scored on: Breaking rules found, Tickets mapped to causes, Correct replacements, Graceful degradation.

Working from: summary_css.md, device_tickets.md.

3. state flowcore

Map the scheduler state machine that ends double-booking

Bluebell had 41 double-booked appointment slots last month, and clinicians are furious. The scheduler's states are implicit: booleans scattered across components, nothing modeling the in-flight request. The team agreed to design an explicit state machine before touching more code, and you own the design.

You deliver: An explicit state machine design for the scheduler with a transition table and a bug-to-gap mapping.

Scored on: In-flight state exists, All three bugs mapped, Complete transition table, User-visible definition.

Working from: scheduler_bugs.md, current_flow.md.

4. performancecore

Cut the booking page load time in half

Bluebell's booking page takes 6.1 seconds to reach largest contentful paint on a mid-range phone over 4G, and analytics shows patients abandoning before the clinician list renders. Marketing is buying ads that land on this exact page, so every second is money. The target from leadership: LCP under 2.5 seconds.

You deliver: A ranked performance plan with per-fix savings estimates that credibly reach LCP under 2.5 seconds.

Scored on: Targets the measured costs, Savings math, LCP-correct focus, Realistic quick wins.

Working from: perf_audit.md.

5. test coveragestretch

Aim Bluebell's test budget where the incidents actually happened

After three patient-facing incidents in one quarter, Bluebell's leadership approved two weeks of dedicated test-writing time, and asked a fair question: why did our 71 percent overall coverage number not prevent any of them? You own the answer and the plan.

You deliver: A test strategy that redirects effort to the incident-bearing modules with specific regression cases and explicit non-goals.

Scored on: Mismatch quantified, Incident-derived cases, Prioritized, typed plan, Explicit non-goals.

Working from: coverage_report.csv, incident_log.md.

How the scoring works

Each deliverable is graded against the rubric written for that mission. Separately, every mission on every path is graded on how you used AI, against the same four criteria:

  • Understood the task. The learner framed the goal for the assistant clearly instead of pasting the brief and hoping.
  • Grounded in the material. The learner directed the assistant into the provided files and based the work on them, not on invented facts.
  • Verified the output. The learner checked claims, numbers, or coverage against the source material before submitting.
  • Iterated with judgment. The learner refined weak parts of the draft with specific follow-ups rather than accepting the first answer.

Both scores, with the work behind them, go on your proof profile. That is what makes a claim like "I can use AI for frontend engineering" something an employer can check.