Leading people. Building coverage. Making risk visible.
I’m Andrés Felipe Celis Gómez — a QA engineer and former QA Lead who combines hands-on test automation with technical leadership.
My experience spans financial services and entertainment across web, mobile, APIs and microservices. I’ve led and mentored QA teams across multiple regions while staying hands-on with API, UI and Java unit-test automation. I also develop AI projects and explore how AI-assisted workflows can strengthen software quality.
Across QA roles, independent technology consulting and postgraduate study, my path brings together hands-on engineering, technical leadership and continued learning. Select a card for details.
An AI-augmented quality loop
This proposed approach connects software development, QA and evaluation of generative and agentic AI. Bounded agents can accelerate analysis and investigation; people define quality, verify failures and own release decisions.
AI for QA Draft cases, test data and investigation hypotheses for human review.
QA for AI Evaluate variable outputs, multi-step agent actions, tool use and safety.
01
Specify & assess risk
Map user journeys, architecture and acceptance criteria. Include AI and agent failure modes when they are part of the product.
Agent skill
Specification review · risk mapping
02
Build & design coverage
Pair unit, API, UI and regression checks with reference cases, scoring rubrics and multi-step tasks for AI features.
Agent skill
Candidate cases · synthetic data
03
Execute & evaluate
Run automation and exploratory tests. Inspect outputs, traces and tool calls; probe prompt injection and data exposure where relevant.
Agent skill
Trace triage · failure clustering
04
Release & learn
Review coverage and residual risk for a human go/no-go decision. Turn defects and live signals into new regression cases.
Agent skill
Evidence summary · case updates
AI feedback loopObserve failures → propose new cases and safeguards → review → re-evaluate.
Recent AI builds
These are software-development projects that reflect curiosity and hands-on learning. They complement my professional QA record; they are not presented as client QA work.
Hardcore AI + 30X - Program DemoJul 2026
Tinwa
An AI assistant for architecture studios, designed to consolidate supplier quotes into proposals that require human approval.
I identified the problem, specified the architecture using AIDLC, implemented part of the planned MVP, orchestrated agents with OpenSymphony and deployed an initial version to production.
AIDLCAgent orchestrationHuman approval
AI Tinkerers + OpenAI HackathonSep 2026
Xentinela
A proposed mobile AI agent to detect signs of phone scams and alert users.
At an AI Tinkerers hackathon sponsored by OpenAI, I helped frame and specify the problem and implemented part of the MVP backend.
Hardcore AI included dedicated QA and security modules. I also completed the MIT Entrepreneurship Online Bootcamp (2018). Earlier, Happtic—a proposed mobile and vibrotactile learning aid—was selected as an Ayudapps winner by Colciencias.