Brandon Taylor LinkedIn

Applied AI & service reliability

Brandon TaylorAI for incident
resolution.

I’m Brandon Taylor. I build AI that investigates failures and recommends a resolution. As an application support leader, I use that evidence to fix the code or process responsible and keep repeat incidents out of the queue.

Recommended resolution
3–5 minutes
From incident evidence to a recommended action in systems I built.
Diagnostic inputs
4 sources
Logs, runbooks, documentation and previous incidents.
Engineering handoff
RCA PR
Automated root cause write-up and a proposed code change for review.

How I apply ITIL

Every incident gets a
problem follow-up.

Restoring service closes the immediate interruption. I link every incident to a known problem or open one with an owner. That gives us a place to establish the root cause, act on it and check whether incidents stop recurring.

Illustrative investigation

A scheduled report misses its deadline.

Choose a stage to trace the decision

Incident management

Get the report to its users.

I start with who is blocked and what they need. The diagnostic system checks the execution log, runbook, documentation and earlier incidents. Before recommending a replay, it checks whether the report was generated but never delivered.

Evidence
The schedule fired. The destination has no report. The last successful run gives us a comparison.
Recovery check
Verify the expected output reached its users. Attach the evidence to the problem follow-up.

Problem management

Make the cause testable.

I compare the failed run with a successful one. Where did they first diverge? An existing problem can collect many related incidents. If the cause is unclear, the problem stays open with an owner and a specific next check.

Evidence
Compare the job timeline, recent repository changes and the handoff that releases the report.
Decision
A retry that never writes a file points toward the code. A finished file with no release owner points toward the handoff. Test the explanation against the timeline.

Technical corrective action

Reproduce the defect in the repository.

In this version of the case, a retry records success before the report is written. A regression test reproduces the missing output. AI can document the RCA and draft a pull request that moves the success check after the write.

Proposed PR · illustrative pseudocode
− mark_complete(run)  write_report(output)+ mark_complete(run)
Repository evidence
The failing test, the relevant code diff and the incident log all point to the same ordering error.
Change control
Review the PR, test results, risk and rollback plan. Obtain change approval, release the fix, then verify delivery on subsequent runs.

Process corrective action

Fix the handoff that nobody owned.

In this version, the report ran successfully and waited for release. Two teams each expected the other to send it. I work with both teams to revise the runbook and rehearse the handoff before the next deadline.

Revised release procedure
BeforeSend the report to a shared queue.
AfterA named owner accepts delivery. A backup takes over at the cutoff.
Workflow evidence
The completed report existed on time. The ticket moved between queues without an owner accepting it.
Process check
Confirm the next handoff is accepted on time. Track reassignments and minutes spent triaging. Route any production configuration change through change control.

Continual improvement

Check whether the work actually went away.

I compare missed deliveries per 100 scheduled runs before and after the fix. I check triage cost per incident and investigation cost per resolved problem, then confirm users are receiving reports on time.

Outcome evidence
Use comparable observation windows, delivery confirmations, repeat tickets and recorded analyst time.
Close the loop
Record the result and update the diagnostic knowledge. If the same failure returns, reopen the problem with the new evidence.

Support economics

What does the fix save?

Fewer incidents, shorter triage and less time on repeat investigations each affect the return. Change the illustrative assumptions to see how much support capacity a fix could free up.

Illustrative support costs

Monthly workload

BeforeAfter
Related incidents tickets per month
Incident triage minutes per ticket
Problem investigations additional hours per month
Labor rate, costs & calculation

Monthly effort equals ticket count × triage minutes ÷ 60, plus separate problem-investigation hours. Year-one benefit values 12 months of recovered hours at the labor rate. Net value subtracts implementation and 12 months of running costs. ROI divides net value by those costs.

Illustrative projection

16hours

Monthly support capacity recovered

Before
32 h
After
16 h
Year-one net value
$5,900
Year-one ROI
69%

Estimates the labor value of recovered time after costs. Actual returns depend on how that time is used.

Selected work

Systems I’ve built.

Leadership in practice

Experience.

2026 / PRESENT

Family Dollar

Application Support Manager, Supply Chain and Enterprise Data Warehouse (EDW)

I lead support across applications that connect merchandising, distribution and enterprise data. I help teams trace a business problem across those boundaries and agree on who owns the next step.

2025 / 2026

Bridgewater Associates

Site Reliability Engineering (SRE) Lead

I built automated incident diagnostics for application support. The systems checked logs against runbooks, documentation and previous incidents, recommended a resolution, and could draft a root cause analysis and a pull request proposing a fix.

2022 / 2025

PeopleReady · TrueBlue

Senior Application Support Manager

I managed Tier 2 support for staffing applications, including JobStack, and built a support copilot to help colleagues work through operational questions.

Skills index

Technology stack.

The tools I use to build and evaluate AI systems, alongside the enterprise platforms my teams support.

Personal engineering

AI systems & development

  • OpenAI Codex
  • OpenCode
  • Google Antigravity
  • OpenRouter
  • Open WebUI
  • Local large language models (LLMs)
  • Multi-agent orchestration
  • Model Context Protocol (MCP)
  • Agent evaluation
Personal engineering

Automation & infrastructure

  • Python
  • Microsoft Power Automate
  • n8n
  • Docker
  • PostgreSQL
  • Redis
  • MinIO
  • Git
  • Gitea
  • Linux
  • Windows Subsystem for Linux (WSL)
  • Cloudflare Tunnel
  • Portainer
Enterprise support

Service operations

  • IT service management (ITSM)
  • ITIL practices
  • Automated incident diagnostics
  • Tier 2 / L2 application support
  • Zendesk
  • Freshservice
  • Jira
  • Major incident management
  • Incident command
  • Root cause analysis (RCA)
  • Problem management
  • Change and release management
Enterprise support

Supply chain & integration

  • Oracle Retail Merchandising System (Retek RMS)
  • IBM Sterling B2B Integrator
  • Electronic data interchange (EDI)
  • Warehouse management systems (WMS)
  • Catalyst WMS
  • Oracle Transportation Management (OTM)
  • IBM i (AS/400)
Enterprise support

Analytics & data platforms

  • Microsoft Power BI
  • MicroStrategy
  • SAP BusinessObjects (BOBJ)
  • Snowflake
  • Microsoft SQL Server
  • Azure Synapse Analytics
  • SAP Business Warehouse (BW)
Enterprise support

Data integration & ETL

  • SAP Data Services (BODS)
  • SQL Server Integration Services (SSIS)
  • IBM DataStage
  • Enterprise data warehousing (EDW)
  • Extract, transform, load (ETL)

Brandon Taylor · Virginia

Let’s talk systems.

Connect on LinkedIn