• Our process

    Five steps from idea to a product that keeps improving.

  • Engagement models

    Spec & quote projects, an agile retainer or team augmentation.

  • Our tools

    The playbook, CMS and project tools we built and use every day.

  • Company

    A boutique digital agency in Cedar Grove, NJ, since 2001.

  • What We Believe

    What we believe and how it shapes the way we work.

  • Team

    Meet the designers, developers and strategists.

  • Testimonials

    What our clients say about working with us.

  • Insights

    Ideas on software, AI and agents from our team.

All case studies Skills-Based Job Matching · Nonprofit

Matching people to jobs on what they can do, not the keywords they used

A workforce nonprofit helps people without traditional credentials get hired for what they can actually do. It needed one platform to replace a patchwork of tools — and to make skills-first matching work at scale.

Industry
Nonprofit / Workforce
Engagement
Product design & build
Cloud
AWS
Users
Job seekers, staff & employer partners

Services provided

  • Product Design
  • Agentic Application
  • Skills-Based Matching
  • Explainable Scoring
  • AWS Deployment

The challenge

Degrees and keywords are poor proxies for ability, and they quietly filter out exactly the people the organization exists to serve. Its staff were stitching hiring “sprints” together across outside platforms, spreadsheets and trackers.

What we built

We designed and built one self-contained platform for job seekers, staff and employer partners. Five agents each own one job — onboarding conversations, résumé understanding, job analysis, sourcing and employer research — and share their work through the platform rather than calling each other.

Just as deliberately, we kept some decisions out of the AI. Segmentation and hard filters are plain, auditable rules; the AI scores only what genuinely needs judgment.

How it works

  1. 01

    Onboard

    A conversational assistant helps job seekers sign up, and their résumé is turned into structured skills — with every AI suggestion approved by the candidate, section by section.

  2. 02

    Understand the role

    Each job is analyzed for skills, and a candidate persona and evaluation rubric are generated before any matching runs.

  3. 03

    Filter fairly

    Hard requirements and distance are applied as transparent rules, not model guesses.

  4. 04

    Match on skills

    Candidates are matched on weighted skill overlap using an industry-standard skills taxonomy — adjusted for recency and confidence.

  5. 05

    Score & explain

    The AI scores technical fit, culture fit and the rubric, and records a verdict, evidence and rationale for every criterion — so “why this job?” always has an answer.

Jenna Payette, Director of Client Services at NP Group
Guardrails · 4 built in

Where people stay in charge

  • Agent results are recommendations, never final decisions — staff approve candidates at a review stage in every sprint.

  • Matching runs are started by staff, by design.

  • Candidates can opt out of AI scoring, and consent is checked before any AI processing.

  • Every match stores its scoring-policy version, so results can be reproduced and audited.

John, NP Group
Results · 3 outcomes

What changed

  • One platform replacing outside tools, trackers and spreadsheets.

  • Every match explained, criterion by criterion.

  • Model choice per agent — OpenAI, Anthropic Claude or AWS Bedrock — with automatic fallbacks.

Technology

Built with

  • AWS
  • OpenAI
  • Anthropic
  • Laravel
  • PHP
  • Python
  • FastAPI
  • React
Pete, CEO and founder of NP Group

Pete CEO & Founder

Have a workflow like this? Let’s prove it with code.

Tell us the process that costs your team the most. In a few weeks we’ll show you, with working software against your own data, what agents and good design can do with it.