KBKevin Bartolome
Mission Control · Personal project, 2026 to present›

AI Engineer · Forward Deployed Engineer · Lead Software QA Engineer · Software Developer

Kevin Bartolome

I build AI agents that investigate, fix, create and review end-to-end (E2E) tests, and I keep a person on every decision that is hard to undo.

US Citizen · Remote or relocation

Read the story in three beatsSee it runningThe BridgeInvestigationsResume
300-500
failing E2E tests cleared a day on one large AI product, in about four hours1
10x
as efficient as investigating failures by hand2
50
fully vetted E2E tests in one day, against 10 a week expected3
20
client investigations at once, about 800 failures in all4

The story, in three beats

Problem, what I built, what changed.

Beat 1 · The problem

Thousands of failing E2E tests a day, and new ones written slowly, all by hand.

I lead a team that investigates 5,000+ E2E test failures a day for about 200 customers.5 Every failure meant someone reading its logs, its trace and the live page to find the cause, then deciding: is the E2E test wrong, or is the product broken? New E2E tests were slow too: the company expected about 10 a week.3 And every E2E test, new or a teammate's, needed a person to check that it does what its steps say.

5,000+E2E test failures a day, for about 200 customers5
✗bil-19 · checkout total is emptyread by hand
✗srch-02 · results in the wrong orderread by hand
✗onb-05 · upload never finishesread by hand
… and hundreds more before noon

Beat 2 · What I built

Mission Control: agents that investigate, create and review E2E tests, and a person on every call that is hard to undo.

A personal project, 2026 to present.6 Three kinds of agent work share one evidence gate and one person on the calls. A lead agent session owns each client, and worker sessions each own one piece of work end to end.4

  1. The workInvestigate Failures grouped by cause on the Investigations page, each root cause with evidenceCreate New E2E tests, every written step mapped to an assertion seen passingReview Each E2E test checked against its steps; a second agent challenges every verdict
  2. AgentsEach kind of work has its own agents, and every claim carries its evidence
  3. Lead per clientWorkers each own one piece, end to end
  4. Evidence gate"Ready" needs a clean local run plus a platform run, same code
  5. A personApproves only what is hard to undo

↻Every correction I make becomes a written rule, and where it can, a check that fails on a repeat.

Beat 3 · What changed

The reading and the writing moved to the agents. The calls stayed with a person.

1 key
for a bug report, not 20 to 30 minutes2
Gated
no agent certifies its own fix: "ready" needs a run id, and guardrails refuse destructive platform actions at the tool boundary7
~1k
tokens per filtered report, down from 3-8k per raw page snapshot8
Vetted
every written step maps to an assertion seen passing, and a second agent challenges every review verdict3

See it running

The real pages, on invented data.

A preview of each page is below. Open one to click through the real page, drawn by its own code from invented records. Each opens on a normal day; the switch at its top shows a busy one.

What needs me right now?

The one page I watch: what needs a person, what each agent session is doing, and what got done.

Open The BridgeSee a busy day

What happened to this morning's failures?

Investigations: hundreds of failures at once, grouped by cause, worked in parallel, each fix proven before it reaches me.2

Open InvestigationsSee a busy day

The full story · about 5 minutes

How I handed most of the E2E test work to AI agents, and what still needs a person

The constraints, the decisions and what each one cost, how it runs, how it stays honest, and one day start to finish on invented data.

Read the full story
  • multi-agent orchestration
  • evidence gates
  • guardrails
  • context engineering
  • MCP
  • Claude Code

About me

Lead Software QA Engineer now, Software Developer before.

The resume, readable here or as a one-page PDF.

View the resumeDownload the PDF

Sources

Every figure on this page is on my resume.

Each footnote lands here. Each line names where its figure sits on the resume and opens that line.

  1. 1300-500 failing E2E tests cleared a day on one large AI product, in about four hoursResume · Lead Software QA Engineer · bullet 2See it on the resume
  2. 210x as efficient as by hand, hundreds of failures at once, and a bug report in one key, not 20 to 30 minutesResume · Projects, Mission Control · bullet 1See it on the resume
  3. 350 fully vetted E2E tests in one day, against 10 a week expected, every written step mapped to an assertion seen passing, and a second agent challenging every review verdictResume · Projects, Mission Control · bullet 2See it on the resume
  4. 420 client investigations at once, about 800 failures in allResume · Projects, Mission Control · bullet 3See it on the resume
  5. 55,000+ E2E test failures a day, for about 200 customersResume · Lead Software QA Engineer · bullet 1See it on the resume
  6. 6A personal project, 2026 to presentResume · Projects, Mission Control · its datesSee it on the resume
  7. 7No agent certifies its own fix, and guardrails refuse destructive platform actionsResume · Projects, Mission Control · bullet 4See it on the resume
  8. 8~1k tokens per filtered report, down from 3-8kResume · Projects, Mission Control · bullet 5See it on the resume