Hiring
Runs hiring on structured scorecards instead of opinions, finds interview slots across a panel’s real calendars, and stops candidates going silent.
Build effort, not elapsed calendar time. Engagements ran alongside study and employment, so this figure is the work itself rather than the span it sat in.
Interface built in code for this case study — not a client screen capture.
01Business problem
Hiring at most companies runs on a spreadsheet and a shared inbox, which is why good candidates go silent for two weeks and interview decisions come down to whoever spoke last.
Who feels it
Applications arrive by email. Someone forwards promising ones to a hiring manager, who replies when they get to it. Interview scheduling becomes a chain of messages trying to find ninety minutes that three people share. Candidates in the middle of the pipeline are forgotten because nothing tells anyone they are waiting.
Evaluation is the weaker half. Interviewers arrive without knowing what they are assessing, notes are written in a document nobody reads, and the debrief is dominated by the most senior or most recent opinion. When feedback is shared before everyone submits, the first strong view anchors the rest — a well-documented effect that a shared spreadsheet actively encourages.
And nobody can answer basic questions afterwards: how long a role took to fill, where the candidates who succeeded came from, or which stage everyone drops out of. The data existed; it was never structured.
02Product overview
An applicant tracking system where each role has a defined pipeline with entry requirements, interviews are scheduled by intersecting real panel availability, scorecards are structured per interview and hidden until submitted, and every stage has a response deadline.
Engineering notes
The pipeline is a state machine per role. A stage advance requires its artefacts — a screening stage requires recorded notes, an offer stage requires an approved compensation band — so a candidate cannot arrive at an offer without the evaluations that justify it. Every candidate at every stage has a response deadline, and the dashboard is built around what has breached it.
Scheduling is the feature that saves the most time. Panel availability is read from connected calendars, converted into free intervals, intersected to find windows long enough for the interview, and filtered against working hours in each interviewer’s own timezone plus a buffer between meetings. The recruiter picks from real options instead of proposing times and waiting.
Scorecards are structured against the role’s defined attributes with a rating and required evidence per attribute. Submissions are hidden from other interviewers until they submit their own, which removes the anchoring effect that makes a shared document a worse instrument than no document. Aggregation shows agreement and disagreement rather than collapsing to an average.
Candidates are deduplicated across roles, so a person who applied last year appears with their history rather than as a new applicant, including any do-not-proceed decision and its reason.
03Key features
10 capabilities, of which 1 are model-backed. The rest are ordinary deterministic software, and the distinction is marked so the AI claim stays honest.
Each role defines its stages and what must exist before a candidate can advance, enforced on the transition.
Every candidate at every stage has a deadline, and the dashboard leads with what has breached it rather than with totals.
Free intervals from connected calendars are intersected to produce real slots, respecting each interviewer’s working hours and timezone.
Structured ratings with required evidence per attribute, hidden from other interviewers until submitted.
Aggregated scorecards show where the panel agrees and where it splits, instead of averaging disagreement away.
Applicants are matched across roles and time, so previous history and prior decisions surface immediately.
Contact details, roles, dates and skills are extracted into fields for search, with the original document always primary.
Offer creation requires approval against a band by amount and level, recorded with the approver and the reason.
Templated updates at each stage including rejections, so candidates are not left silent by default.
Time in stage, drop-off by stage and outcome by source, computed from the pipeline rather than reconstructed.
04User flow
The path from the trigger to the finished record, with each step attributed to whoever performs it — a person or the system.
05Architecture
The processing path first, then the layers it runs on, then the constraints that shaped both.
Path through the system
Application captured, resume parsed, candidate deduplicated.
Requirements checked, stage transition validated against artefacts.
Calendar free/busy intersected into valid slots with buffers.
Blind scorecards collected per interviewer against role attributes.
Aggregated view showing agreement, then a recorded decision.
Band approval, offer issued, outcome recorded to the funnel.
Layers
Why it is shaped this way
06Technical decisions
Every decision here was contested by a reasonable alternative. The trade-off column is the part usually left out.
Why
The scheduling round trip is the single largest source of delay in hiring. Deriving real slots from calendar free/busy turns a multi-day exchange into a selection.
Trade-off
Requires calendar access and correct handling of all-day events, tentative responses and recurring meetings. Interviewers without a connected calendar fall back to declared availability windows.
Why
Anchoring is real and a shared feedback document makes it worse. Independent submission before disclosure produces genuinely independent signals, which is the only reason to run a panel.
Trade-off
A slow interviewer blocks the debrief view, so a deadline plus an explicit override — recorded, with a reason — releases the aggregate.
Why
A number without a reason is not reviewable, and a hiring decision that cannot be reviewed cannot be improved or defended. Requiring evidence also improves the ratings themselves.
Trade-off
More friction for interviewers. Mitigated by keeping attributes few and specific rather than a long generic rubric.
Why
A pipeline whose stages mean nothing produces reporting that means nothing. Making the requirement structural keeps the funnel data honest.
Trade-off
Genuine exceptions exist — a referral hired fast — so an override path exists, logged and visible in reporting rather than silent.
Why
Parsing is imperfect and a wrongly parsed date can eliminate a candidate. The structured fields power search; the original is what a human reads and the parsed values are editable.
Trade-off
Search can miss a candidate whose resume parsed poorly, which is why keyword search runs over the raw text as well as the fields.
07Challenges
Not the setup work. These are the problems where the first implementation was wrong and had to be reconsidered.
Approach
Busy periods are merged into a normalized set per interviewer, inverted into free intervals over the search horizon, then intersected across the panel. Each candidate window is filtered against that interviewer’s working hours in their own timezone, a configurable buffer before and after existing meetings, and a minimum notice period. Results are ranked by how few interruptions they cause to already-busy days.
Outcome
Recruiters pick from a handful of slots that all actually work, and the back-and-forth disappeared from the process.
Approach
Candidate identity is matched on normalized email and phone, with name plus employment-history overlap as a weaker signal that routes to review rather than merging. A merged candidate keeps all applications and any prior decision with its reason attached.
Outcome
A previous do-not-proceed decision surfaces before another panel spends five hours rediscovering it.
Approach
All storage and computation is in UTC with the interval logic timezone-free. Working hours and rendering are applied per participant using their zone, and daylight-saving transitions are handled by resolving working hours against the local calendar date rather than a fixed offset.
Outcome
Invitations land in everyone’s working day, including across a daylight-saving change, which was where the naive offset arithmetic broke.
08Business value
Stated qualitatively on purpose. Invented percentages are the easiest thing to put on a portfolio and the easiest thing to see through.
Operational effect
What would change at scale
09Interface
Dense operational views rather than dashboards. These are used for hours at a time, so the priorities are legibility, keyboard flow and state that is never ambiguous.
Recruiter dashboard
Candidates past their stage deadline first, then interviews today, then awaiting scorecards.
Slot picker
Real intersected panel availability with each interviewer’s local time shown per option.
Scorecard
Role attributes with a rating and a required evidence field, and other submissions hidden until this one is in.
Agreement view
All scorecards side by side with agreement and split highlighted per attribute.
The screen above is built in HTML and CSS for this case study. It reproduces the layout, states and vocabulary of the real build without exposing client data, which is why it exists rather than a screenshot. It is evidence of design and of the decisions behind it. It is not a photograph of a deployed system, and no part of it is a capture of anyone else’s product.
10Technologies
Chosen for the shape of the problem, not for novelty. Anything unusual is justified in the decisions section above.
11Technical preview
There is no repository link on this site. What is available instead is the module structure, the data model, the interface contracts and annotated excerpts written for this case study — enough for a technical reviewer to judge the engineering.
Technical preview
Selected implementation details are available for technical review: component structure, data models, architecture notes and sanitized code excerpts. Some client-specific source material stays private because of confidentiality. Enter the project access keyword if you have been given one — the same one works across every case study.
A presentation convenience, not authentication. This is a static site, so the keyword ships to your browser with the page and anyone reading the bundle can find it — worth saying plainly rather than dressing it up. Nothing confidential is stored behind it.
Inside this section
For private client work, contact me for a walkthrough.
Next step
Send the process, the constraints and the deadline. You will get an honest scope, an architecture sketch and a timeline before any commitment.