load-test scenario
Documented performance-engineering write-up.
Performance engineering · Backend systems · AI delivery
I make complex systems measurable, fast, and shippable.
Lead Performance Engineer and product builder. I trace bottlenecks across the stack, turn architecture into working software, and direct AI agents through bounded, reviewable delivery.
Open to senior engineering roles and focused consulting engagements.
Method
A system is a set of decisions you can inspect.
Change the lens, inspect a node, and follow how human direction becomes verified output.
Human direction defines the work; isolated execution creates evidence for review.
Human direction
Intent before automation
A person sets the objective, scope, constraints, and stop conditions.
1/5
Evidence
Evidence, with context
Signals from documented work—not a wall of vanity numbers.
in the AgentOS workflow model
A work-in-progress orchestration system.
in Vitrina's production Node API
The public site is available; the production WhatsApp channel is paused.
Selected work
Six systems stories
Selected for range: scale, constraints, product delivery, orchestration, and open tooling.
Performance engineering practice
Performance at Scale
Load and observability work focused on finding constraints before they become customer incidents.
Read case studyPublic site available; production WhatsApp channel paused
Vitrina
A WhatsApp-native website generator for local Mexican businesses, with a zero-dependency Node backend.
Read case studyIn development
MISO OS
An in-development point-of-sale system whose FastAPI service is the product backend; Go compiles and validates contracts.
Read case studyIn active development
AgentOS
A work-in-progress, agent-agnostic operating system for human-directed product delivery.
Read case studyProduct portfolio case study
MX Stock Analyzer
A bilingual Mexican-equities analysis product built on reusable open-source market-data tooling.
Read case studyOpen-source tools
Open-source Performance Toolkit
Open tooling for generating load tests, running benchmarks, and making performance evidence easier to inspect.
Read case studyMethod
AI changes throughput. It does not own the decision.
Claude and Codex operate inside a human-directed engineering system with explicit roles, context, isolation, review, verification, and authorization.
A person assigns the outcome. Claude plans or reviews; Codex executes an approved, bounded task. Neither silently inherits authority.
Services
Focused interventions
Useful when you need a decision, a performance baseline, a safer agent workflow, or a bounded product slice.
01Fixed price · 3–5 working days
Systems diagnosis
A focused technical readout of the constraint, evidence, and next experiment.
InvestmentUSD 500MXN 10,000
Credited toward a Performance audit booked within 30 days.
Get a quote02Fixed price · 2–3 weeks
Performance audit
A load, latency, and observability review with a prioritized remediation plan.
InvestmentUSD 1,500MXN 30,000
Get a quote03Fixed price · 3–4 weeks
Agent workflow
A human-directed AI workflow designed around roles, controls, and verification.
InvestmentUSD 2,500MXN 50,000
Get a quoteExperience
Systems work across the stack
Backend foundations became performance practice, then technical leadership.
- 01
Lead Performance Engineer
EPAM Systems
- Lead performance engineering for backend and distributed systems in cloud environments, designing testing strategies and establishing CI/CD quality gates to reduce production risk
- Integrate observability tooling and collaborate with engineers and technical leaders to identify bottlenecks, plan capacity, and build repeatable performance practices for APIs and microservices
- 02
Software Engineer
SailPoint Technologies
- Developed and executed performance tests in Kubernetes hosted in AWS EC2 services using k6, saving $100k in resource allocation and improving response performance from 10s to 50ms
- Exported logs to AWS S3 buckets using AWS CloudWatch to monitor performance and availability from PostgreSQL databases, reducing downtime by 40%
- 03
Software Engineer
Globant Consulting
- Conducted cloud load testing with k6 framework for services up to 100k RPS and implemented CI/CD asynchronous pipelines for an online dealer auction platform
- Analyzed New Relic results to provide optimal resource management for Kubernetes across multiple services, improving performance and cost optimization
Writing
Notes from the field
How the measurements, failures, and design constraints change the system.
Shipping a SaaS With Zero node_modules
What I learned building Vitrina — a WhatsApp-native website generator for local Mexican businesses — on a hand-written, zero-dependency Node backend, and why Meta nearly killed it overnight.
An Operating System for AI Agents
AgentOS runs a full org chart of role-based AI agents to take a product from a context pack to a deployed MVP. Notes from building it — and why the interesting problems are boring infrastructure problems.
Generating Load Tests From OpenAPI With LLMs
k6gen turns an OpenAPI spec into runnable k6 load tests using a deterministic core and a tightly-constrained LLM pass. Here's why the boring part matters more than the AI.
Bring the system that refuses to explain itself.
For engineering roles, consulting, or a precise technical collaboration, send the constraint and the evidence you already have.