0%

Arca

Description

Arca is workforce management for nursing homes. I led product and design end to end, from the MVP scope to every screen, and took it from first commit to a complete platform in 6 months, with an AI copilot at its core.

Impact

6 months

From first commit to a complete product

Team of 3

I led every product and design call

1 AI copilot

Catches call-offs and overtime before the shift starts

ProjectWeb platform + worker app
Year2026
RoleProduct & design lead
LocationUS (remote)

Problem

Nursing homes run on call-offs. A 6am call-off means an hour of phone calls, overtime nobody sees until payroll, and a federal staffing report rebuilt by hand every quarter. Most tools record the schedule. None of them work it.

Solution

One platform for the whole shift: plan it, staff it, clock it, report it. And an AI that watches the schedule and brings the fix before anyone has to ask.

Arca is built and ready, but not launched yet. There are no customer numbers here: this is how it was built, and why each call was made.

AI copilot

SmartOps

An AI that works the schedule, not one that answers questions about it.

  • Scans every department every 10 minutes for five problems: call-offs, overtime, open shifts, understaffing and overstaffing.
  • Each one comes with a worked-out fix and the top 3 people who can cover without going into overtime.
  • Reads the team chat. A worker texts “can’t make it tonight”, the AI finds the call-off and the shift, and a cover card appears in the thread.
  • Asks up to 25 people at once, and can give the shift to the first one who says yes.
  • Moves people between units to balance the floor, asks them first when their day or shift changes, and applies the plan only when everyone agrees.
Live preview
SmartOps

Cover is found and asked before you reach for a phone.

Overtime is caught before the shift is worked.

Tonight's gap never hides behind next week's.

A short day filled from a heavy one. No agency call.

Why it suggests and never acts alone

Schedulers own the schedule and answer for it. An AI that rewrites a published schedule loses their trust the first time it’s wrong. SmartOps prepares every fix; one click from the scheduler applies it.

Why I cut chain moves

Real floors move people in chains: A covers B, B covers C. I designed for it, then capped moves at 3 and blocked any that leave another unit short. A plan a scheduler can read in five seconds beats a clever one they won’t trust.

SmartOps staffing tab: where a unit needs someone, who to assign, and who to move off an overstaffed shift

Scheduling

  • Staffing targets by unit, shift and position, so a gap shows before the week is published.
  • Master schedules and rotations fill the weeks ahead without touching any shift the scheduler edited by hand.
  • Open-shift rules decide who sees an open shift, and when.
  • Agency cover with bill rates by agency and position, checked against the agency's own file.
Why a review queue instead of auto-fill

A rotation that silently overwrites a hand-edited shift breaks trust in the whole grid. Every projected shift waits for review, and edited ones are never touched.

Live preview

Time & attendance

  • Clock in from the phone inside a 150 m geofence, or from a wall tablet with a 4-digit PIN.
  • Grace, rounding, breaks, missed clock-outs and duplicate cards handled by rules, not by hand.
  • "OT Creep" flags a shift running past its window before it turns into overtime.
  • Payroll export with hours per worker: regular, overtime and time off.
Why automatic break deductions start off

A default that cuts someone's pay by mistake is a wage problem, not a UX problem. The facility turns it on knowingly.

Arca timecards for the pay period, with late punches and missing timecards flagged

Chat

  • One workspace per company and a channel per facility, with membership driven by the roster.
  • Letting someone go removes their access; their messages stay on record.
Why chat lives inside the product

Chat is where call-offs actually arrive. Keeping it inside Arca is what lets SmartOps hear them.

Arca chat: a worker calls off and SmartOps suggests who can cover, right in the conversation

PBJ

  • Builds the quarterly federal staffing report (PBJ) straight from timecards.
  • 59 rules: federal meal deductions, hours split at midnight, a permanent ID per person, and the deadline 45 days after the quarter.
  • The file is checked against the official format before it's saved, and every issue names who can fix it.
Why it never says "compliant"

Compliance is decided by the government, not by software. Arca says what's ready and what blocks the submission, and stops there.

Arca PBJ report ready to submit: the quarter, totals, the generated file and the steps to send it to CMS

What I left out

  • Time-off plans with carryover, caps and tenure tiers. Every facility writes these differently, so a flat accrual rate shipped instead.
  • Shift swaps. Specified, not built.
  • Union and multi-tier overtime. One daily and one weekly threshold per facility.
  • An AI chat that edits the schedule. SmartOps decides; the chat only explains.

Conclusion and final thoughts

Arca covers the whole shift, from the schedule to the federal report, and it was built in 6 months because every call was made once and early.

A product built for the scheduler at 6am, and for the worker who never chose it.