Portfolio / Practical AI / MarketSync TA

MarketSync TA: an executive talent intelligence platform.

A working application built to answer a question recruiting leaders get asked constantly and rarely have evidence for: what is actually happening in this labor market, and what should we do differently because of it? MarketSync TA pulls public labor market data into the moment a hiring decision is made — intake, calibration, relocation, pipeline review — and exports the result as a brief an executive can read.

Open the live app See the workflows Private repo — available on request
Role
Sole designer, engineer, and operator
Status
Live public demonstration
Core stack
Next.js · TypeScript · Tailwind · Cloud Run · Firestore
Data
Public labor market APIs · synthetic candidate records
The problem

Recruiting teams are asked for headcount. They are rarely asked for a point of view.

In most organizations the recruiting function receives a requisition, agrees to a target date, and starts working. The market context that would change the conversation — how many competing postings exist for that role in that geography, what the real supply looks like, what a relocation actually costs the candidate — arrives late, arrives inside a vendor deck, or does not arrive at all.

The gap is not data availability. Public labor market data is abundant. The gap is that the data does not live where the decision is made. By the time someone assembles a market read, the intake meeting is over, the offer is out, and the hiring manager has already anchored on a timeline nobody stress-tested.

I have run this problem from the inside — ATS migrations, capacity modeling with a CEO and CFO, pay-scale reviews, hiring manager enablement across multiple organizations. The recurring failure mode is the same: recruiters have the relationship and the instinct, but not a repeatable way to bring evidence into the room.

MarketSync TA is my attempt to build that instrument rather than write another framework about it. It is a portfolio application — a demonstration of how the workflow could work — and it is built to production engineering standards so the architecture argument holds up under scrutiny.

Design principles

Four decisions that shaped the build.

Most recruiting tools optimize for output volume. This one optimizes for whether a hiring manager or an executive will actually trust and act on what comes out of it.

01

Every number carries its provenance.

Each data point returned by the platform is tagged with its source, a confidence indicator, and a fetch timestamp. A market figure an executive cannot trace is a figure they will not use in a board conversation — so traceability is a first-class field, not a footnote.

02

Degraded providers are declared, not hidden.

A Data Source Registry shows each provider as live, sandbox, or degraded. When an upstream API is unavailable, the interface says so rather than quietly serving stale or empty results. Silent failure is how analytics tools lose the credibility they take months to earn.

03

Output has to leave the tool.

Strategy work gets consumed in documents, decks, and ATS fields — not in a dashboard nobody logs into. Intake agendas map to Greenhouse- and Workday-style custom fields, and the Executive Strategy Playbook exports to Markdown in one click so the brief travels to wherever the decision is being written down.

04

No real candidate data, by design.

The Candidate Hygiene Dashboard runs entirely on synthetic sample records. A public portfolio demonstration is the wrong place for real candidate PII, and building it correctly from the start is part of the argument for how AI should enter recruiting workflows at all.

What it does

Seven workflows, one decision path.

The modules are sequenced the way a real requisition moves: read the market, run a better intake, calibrate the offer, verify the pipeline, and hand leadership something they can act on.

Adzuna API

Market Demand & Hiring Volume

Posting volume, demand trends, and geographic heatmaps for a role and market. The opening evidence for whether a timeline is realistic before anyone commits to it.

Market Intel Live provider
AI Playbook

Dynamic Hiring Manager Intake

Generates a structured intake agenda grounded in that role's market read — the questions a recruiter should be asking, and the market realities the manager needs to hear before committing to a plan.

Intake · Enablement Core workflow
ATS Integration

ATS Custom Field Mapping

Maps generated intake output to Greenhouse- and Workday-style custom fields so the strategy survives contact with the system of record instead of dying in a document.

Greenhouse · Workday patterns Core workflow
Markdown export

Executive Strategy Playbook

Compiles the market read, intake plan, and calibration inputs into a single executive brief with one-click Markdown export — built for the leader who wants the argument, not the dashboard.

Executive Communication Core workflow
Teleport API

Relocation Offer Calculator

Converts a relocation conversation into a defensible cost-of-living offset range instead of an instinct-driven negotiation — designed to take guesswork out of geographic offer decisions.

Compensation · Calibration Live provider
Abstract API · synthetic data

Candidate Hygiene Dashboard

Validates pipeline record quality and email deliverability across a synthetic candidate set — surfacing the data problems that quietly distort every funnel metric a recruiting leader reports.

Data Quality Synthetic data
Governance

Data Source Registry

A live status view of every configured provider — Adzuna, Teleport, Abstract, Firestore, BLS, O*NET, USAJOBS — showing whether each is live, sandboxed, or degraded, with source and confidence metadata attached.

Trust · Transparency Always visible
Try it

Walk the workflow yourself.

The application is publicly accessible. Run a market query, generate an intake agenda, and export the brief — the fastest way to judge whether the argument holds.

No login required Open the app →
Architecture

Built the way I would want a vendor to build it.

The engineering choices are deliberate. A recruiting leader evaluating AI tooling should be able to ask how deployment credentials are handled and get a real answer — so I built this one to be able to answer that question about my own work.

Application
Next.js (App Router) · React · TypeScript · Tailwind CSS — server components for data-fetching routes, typed provider contracts so a degraded API surfaces as a handled state rather than a runtime error.
Runtime
Google Cloud Run (us-central1) — scales to zero between sessions, which is the correct cost profile for a public demonstration with bursty traffic.
Containerization
Multi-stage Docker build on node:20-alpine — build dependencies stay out of the runtime image, keeping the deployed artifact small and cold starts short.
Persistence
Cloud Firestore — candidate and system state, with the synthetic dataset kept isolated from any live provider response.
CI/CD
GitHub Actions — build, containerize, and deploy on merge; no manual deploy step and no local credentials in the release path.
Security
Google Workload Identity Federation — keyless deployment. No long-lived service account JSON key is stored in the repository or in CI secrets, which removes the most common credential-leak path in this kind of pipeline.
Data providers
Adzuna (job market) · Teleport (cost of living) · Abstract (email validation) — with routing and infrastructure in place for BLS, O*NET, and USAJOBS.
Honest scope

What this demonstrates — and what it does not claim.

This is a portfolio application, not a production client system. I would rather state the boundaries plainly than let a demo imply results it has not earned.

What it demonstrates

  • Designing a recruiting workflow around evidence and decision points rather than activity volume.
  • Shipping and operating a containerized application on managed cloud infrastructure end to end.
  • Integrating multiple third-party data providers behind a consistent contract with source, confidence, and freshness metadata.
  • Applying secure deployment practice — keyless federated identity, automated release, no stored credentials.
  • Translating labor market data into artifacts an executive audience will actually read.

What it does not claim

  • No measured business outcomes. The workflows are designed to accelerate intake preparation and architected to reduce guesswork — not benchmarked against a control group.
  • All candidate records are synthetic. No real candidate data has been processed by this application.
  • Some providers run in sandbox or degraded mode depending on credential availability; the registry reports current status.
  • It is not a production client system, and it is not currently operating inside any employer's hiring process.

Where it goes next.

  • Complete the public-data layer. Routing exists for BLS, O*NET, and USAJOBS; wiring them fully would broaden the market read beyond posting-volume signals into occupational and wage data.
  • Write back to the ATS. Field mapping is the first step; a Greenhouse API integration would let an intake plan land directly on the requisition.
  • Workspaces and authentication. Required before anything resembling multi-team use, and deliberately out of scope for a public demonstration.
  • Instrumentation. If this ran inside a real function, the honest next step is measuring intake cycle time and offer-decline reasons before and after — so the claims section above could finally include numbers.

Why I built it.

I have spent twenty years inside recruiting functions and the last several building the systems that hold them together — ATS migrations, capacity models, scorecards, enablement programs, and a set of AI workflows used by working recruiters.

MarketSync TA is the version of that work where I owned the whole stack: the problem framing, the product decisions, the architecture, the deployment pipeline, and the security posture. It exists because I think the next generation of recruiting leadership will need to speak both languages — the operating model and the system underneath it.

If you are building in HR technology, workforce intelligence, or B2B SaaS and this is the kind of thinking you want in the room, I am happy to walk through it.

Want the walkthrough, or the code review?

Contact

Building something where recruiting has to be a system, not a service desk?

I am open to remote Head of Talent Acquisition and talent operations roles with technology, HR/workforce tech, and B2B SaaS companies, and to selected consulting engagements. Start with a 20-minute conversation — if it is not a fit, I will say so.

Ryan Borths © 2026 Ryan Borths · Built with intention