I turn product ambiguity intoshipped AI systems, production software, measurable proof or customer value.
Full-stack software engineer in San Francisco. I combine systems engineering, production AI delivery and 17 years of enterprise GTM leadership to build the right thing—and prove it works.
Three products. Three kinds of engineering judgment.
Not a gallery of screenshots. These are short case studies about the tradeoffs, systems and verification behind production-minded software.
01 / 03
AI speaking coach in production
EnglishChunks.guru
A privacy-first learning product that turns open speech models into useful, specific feedback.
Constraint
Give English learners pronunciation feedback that is fast and specific without sending sensitive voice recordings to a third-party AI API.
System
I split the product across a Next.js application and a self-hosted speech pipeline on DigitalOcean. Silero VAD, NVIDIA Parakeet and wav2vec2 handle scoring; Gemma writes personalized feedback; an admin queue keeps content quality human-reviewed.
Outcome
The result is a subscription-ready daily practice loop with level-based lessons, writing feedback, streaks and progress tracking—built around a deliberate privacy boundary.
A financial model designed around how people actually reserve and spend money.
Constraint
Make forecast-versus-actual budgeting understandable while preserving the precision needed for balances, payment buckets and threshold alerts.
System
I modeled each expected payment as its own bucket, then connected the React product to an Express and Prisma backend with authentication, runtime validation and date-safe financial logic.
Outcome
Users can build a budget in guided or complete mode, see live balances and projections, and act on in-product insights across an accessible responsive interface.
A focused converter expanded into a resilient market research workflow.
Constraint
Go beyond a one-off conversion and help users compare, revisit, share and export exchange-rate decisions without making the interface feel like a trading terminal.
System
I built rate history, multi-currency comparison, favorites, shareable pair URLs, keyboard shortcuts, CSV export and an offline fallback on top of live central-bank data.
Outcome
The product balances information density with accessible interaction, backed by broad component and behavior coverage rather than demo-only happy paths.
AI makes implementation faster. That raises the value of framing, architecture, review and verification—the parts of the work I keep firmly human-led.
01
Frame the signal
Clarify the user, business outcome, constraints and evidence that would make the work worth shipping.
Problem brief + success criteria
02
Specify the system
Turn ambiguity into flows, data boundaries, architecture decisions and testable requirements before code.
Constitution + spec + plan
03
Orchestrate the build
Use coding agents for leverage while I control decomposition, integrations, review and the decisions with lasting cost.
Small tasks + reviewed changes
04
Prove the outcome
Verify behavior, accessibility, failure states and production constraints, then document what the evidence says.
Tests + audits + shipped product
Experience
Engineering depth. Commercial context.
I started in telecommunications, spent 17 years helping enterprises adopt complex software, and returned to hands-on engineering with a sharper understanding of customers, adoption and revenue.
2026 — now
Full-Stack Software Engineer
Independent products · San Francisco
Building and deploying full-stack and AI products with modern TypeScript, Python, open models and production integrations.
20+ projects shipped
AI speech systems
Spec-driven delivery
2017 — 2026
Sales leader
Genesys · Colombia and Latin America
Led enterprise CCaaS growth across sales, partners, presales, marketing, customer success and engineering.
$19M ACV pipeline
$4.2M bookings
40+ new customers
2008 — 2016
Enterprise account executive
Microsoft · Colombia
Helped enterprise and public-sector customers adopt Office 365, Azure and unified communications.
Up to 185% of quota
3× Gold Club
Cloud revenue growth
1999 — 2008
Telecom and network engineer
Ericsson, Damovo, Getronics and Telmex
Designed voice and network systems, including a multitenant IP-PBX platform that reached 3,000 seats in one year.
Systems engineering
Solution design
3,000 deployed seats
Elsewhere
Follow the work in progress.
Code, write-ups, challenge solutions and the occasional opinion on building with AI — pick the channel you already use.
I’m looking for a full-stack AI engineering role where product judgment, customer context and rigorous delivery all matter. Tell me what you are trying to ship.
Follow the work in progress.
Code, write-ups, challenge solutions and the occasional opinion on building with AI — pick the channel you already use.