The HR AI Architect
A practitioner's framework for HR + AI

Most HR AI advice is written by people selling something.

This isn't. It's the method I use to bring AI into HR teams safely: built from tools I shipped and piloted with real stakeholders, not from theory.

by Bianca Rohr — HR Data & AI Analyst, AI Solutions Builder

load in  ·  HR reality: real pain, real risk, a lot of hype
01

Pains

Start from the HR problems AI can actually solve, and be honest about the ones it can't. Document fraud in onboarding was one worth solving.

hr-idv-fraud-detection
read the full write-up →
02

Prompts

Production-tested prompt patterns for real HR workflows, with guardrails written into the prompt rather than hoped for.

hr-ai-prompt-library
read the full write-up →
03

Processes

How AI actually reaches a workflow: the integration layer, the rollout, the training, and the governance that earns sign-off.

hris-mcp-connector
hris-offboarding-ticket-automation
read the full write-up →
04

Frameworks

Decision tools that keep a human in the loop: scoring rubrics, bias guardrails, and verify-don't-accuse design.

ai-resume-screener
read the full write-up →
out  ·  AI that leadership, legal, and a DPO will actually approve
How I work

Governance-first, adoption-focused, production over portfolio.

Governance-first. Every system I build ships with human-in-the-loop design, bias guardrails, and documentation a DPO can read cold and sign off on. The restraint is the feature, not a constraint on it.

Adoption-focused. A tool nobody uses solves nothing. I build the training and change management that get a team actually working differently, not just a clever script that sits unused.

Production over portfolio. The work behind this framework ran on real HR workflows with real stakeholders, with the messy edges that only show up once something is live.

Evidence

The framework, as shipped code.

Each pillar above points to something real. Here they all are.

hris-offboarding-ticket-automation n8n A production automation pattern: polls an HRIS for terminations, validates and deduplicates, opens the offboarding ticket, and escalates bad data to a human. Shipped and reviewed.
hris-mcp-connector Python A Model Context Protocol server exposing read-only HRIS tools to AI clients, over a swappable mock data layer.
hr-idv-fraud-detection Python Identity-document fraud signals from EXIF metadata. Signals for human review, never verdicts. Piloted in production.
ai-resume-screener Python Resume fit-scoring with the Claude API, structured rubrics, bias guardrails, and a name-bias invariance test.
resume-integrity-checker Python Flags resume claims for verification, with a neutrality guard enforced in code. Verify, don't accuse.
hr-ai-prompt-library Markdown Production-tested prompts powering an HR team's daily AI workflows, with a governance review checklist.
Connect

Building something at the intersection of HR and AI?

I'm open to AI engineering and HR AI roles. The clearest picture of how I work is the code itself. Start there.

View my work on GitHub →