Brandon Bates · employer-facing technical profile

I build AI workflow systems for high-stakes operations.

I bring 13 years of ICU travel nursing experience into healthcare AI, human-in-the-loop workflow design, and infrastructure-heavy automation. TwoSevens.ai is my public workbench for showing how I think, build, and operate.

Résumé PDF
Code samples available on request
LinkedIn[email protected]
13 years
ICU travel nursing

High-pressure clinical work, handoffs, escalation, and judgment.

AI
workflow implementation

Mapped workflows, structured outputs, review gates, and rollout paths.

HITL
clinical systems

Human-in-the-loop validation for synthetic, non-PHI clinical examples.

~25
Proxmox lab nodes

Lab cluster operation, service separation, and recovery practice.

SQL
data pipelines

PostgreSQL-oriented extraction, normalization, and validation work.

TS
product work

Next.js, React, TypeScript, Tailwind, and product-quality interfaces.

What I build

Systems for work where errors, handoffs, and hidden state matter.

I focus on clinical and operational workflows where AI needs structure, review gates, clear ownership, and reliable handoffs — not black-box automation.

AI workflow automation

Map messy work, define review gates, structure the data, and ship tools that keep humans in control.

  • Workflow maps before model calls
  • Structured outputs and exception paths
  • Auditable review loops

Clinical implementation

Translate bedside judgment into safer healthcare AI workflows with validation thresholds, escalation paths, and non-PHI demos.

  • Clinical risk awareness
  • Human review as a first-class requirement
  • Synthetic data and conservative public boundaries

Lab infrastructure operations

Run the pieces behind the workflow: services, queues, monitoring, deployment practice, and recovery paths.

  • Proxmox lab and service orchestration
  • Service separation and observability
  • Operator runbooks and recovery thinking

Synthetic proof block

Clinical workflow validation, shown with safe sample data.

This screenshot-style block is synthetic. It shows the kind of workflow I build: extraction, validation, flags, and a human review gate before anything moves downstream.

Synthetic clinical validation

No PHI. Sample data only.

Human review

Intake

Synthetic case loaded

Complete

Extraction

Findings structured

Review

Validation

Thresholds checked

Flagged

Human review

Coder gate required

Open

Work examples

Implementation range, labeled honestly.

Clinical AI is the strongest employer signal. Operations and MRO show workflow translation. Trading, security, and music show data scale, boundary-setting, and product range without overclaiming public availability.

Protected demo

Clinical AI workflows

Clinical workflow work built around extraction, normalization, validation, and human review before anything moves downstream.

Employer proof

Shows how bedside judgment translates into structured AI workflows with auditability and explicit safety boundaries.

  • High-stakes workflow mapping
  • Human-in-the-loop review gates
  • Synthetic / non-PHI demo boundaries
Internal proof

Lab infrastructure operations

Distributed systems practice, pipeline reliability, and operator-controlled orchestration in a Proxmox lab setting.

Employer proof

Demonstrates hands-on infrastructure practice behind workflow systems, without implying enterprise production ownership.

  • ~25 Proxmox lab nodes operated
  • Queue-backed orchestration
  • Monitoring and recovery discipline
Prototype

MRO / RegenITAD workflow

AI-assisted asset-listing workflow for ITAD and surplus equipment with extraction, enrichment, normalization, and review.

Employer proof

Shows the same clinical-style review discipline applied to messy operational and asset workflows.

  • Workflow-gated state transitions
  • Evidence-tracked closeout
  • Compliance-aware document generation
Internal-only

Trading / ML research

Simulation-only ML research surface for historical market data, feature engineering, replayable backtests, and risk-control review.

Employer proof

Shows data and systems discipline without exposing financial execution or private infrastructure.

  • 8B+ historical data points
  • Replayable backtest discipline
  • Risk controls before execution
Protected boundary

Security workflow boundary

Controlled security workflow surface focused on defensive monitoring, reporting hygiene, and explicit authorized-use boundaries.

Employer proof

Shows judgment around risk, access control, and public-safety constraints for sensitive tooling.

  • Defensive monitoring posture
  • Evidence and reporting trails
  • No public offensive tooling
Live + planned

Music and SongCraft

Public listening surface plus planned AI-assisted song commerce workflow with generation, review, and delivery gates.

Employer proof

Shows product-surface range while keeping unfinished commerce claims clearly labeled as planned.

  • Public catalog and listen demo
  • AI workflow with human review
  • Commerce path scoped conservatively

Why this matters to employers

This is an implementation profile, not a demo reel.

The useful signal is the combination: clinical judgment from real ICU work, technical delivery across a modern web stack, and enough lab infrastructure experience to reason about operations after the prototype works.

I understand high-stakes work from inside the workflow, not from a whiteboard.

I can explain where AI belongs, where human review belongs, and where automation should stop.

I move between product UI, data pipeline, and lab infrastructure without treating any layer as someone else’s problem.

I use TwoSevens.ai as a public workbench for safe examples, not as a vague agency pitch.

Available for evaluation

Work with me on implementation-heavy AI

I am interested in employer conversations, clinical AI implementation, workflow automation, and infrastructure-heavy roles where operator judgment matters.