A selection of projects across industries — from conversational commerce to AI-powered real estate.
Protected by NDA
Enterprise · Supply-chain AI
NVIDIA — from spreadsheets to a live supply model
Replaced offline spreadsheet guesswork with a live model NVIDIA leadership takes straight into tradeoff conversations. R1 shipped, R2 finalizing — the hard part was making four simultaneous constraints legible on one screen.
TeamPM · PO · Front Eng · QA · AI · Ops users · NVIDIA back‑end
StatusLive · Under NDA
Three releases, one direction
The product grew by shedding assumptions, not adding features.
Each release redefined the ceiling of what the tool could model — and how much of the planning process it could own.
R1
Proved the concept. A single‑constraint model anchored to GPU wafer supply. Replaced offline spreadsheet guesswork with a view leadership could put straight into a tradeoff conversation.
R2
Made it real. Four simultaneous constraints across four product families, plus an agentic onboarding workflow that compressed weeks of manual setup into a single guided session — with human approval at every gate.
R3
Makes it infrastructure. In scoping now. The vision: a platform that self‑onboards new products the moment they appear in the roadmap, runs baseline scenarios overnight, and briefs leadership before anyone asks.
What I designed
Four surfaces around one scenario.
The whole product orbits the scenario as the unit of work. Everything connects back to it.
01
Scenario Manager — overhauled in R2. Faceted filtering, sorting, favorites, bulk actions, cross‑user sharing, and double‑click‑to‑open. The entry surface for every planner session.
02
Bear Chart — the constraint‑driven visualization at the core of the dashboard. Grey bars show the max supply envelope; colored lines overlay each additional constraint. The limiter is readable at a glance. Series toggles, week/month/quarter selectors, and per‑family color management added in R2.
03
Scenario Creator — a dense configuration platform for defining inputs, assumptions, and constraints. Designed to let ops users move fast without losing precision.
04
AI chat — interpretive layer scoped per scenario. The model already has the full scenario context — no copy‑pasting numbers into a generic assistant.
The first version was the most ambitious. The shipped version is the most useful.
01
Reshape, don't protect. Earlier iterations were more robust than what's live today — feedback kept shrinking the surface until what remained was what operators actually needed.
02
Busy users reward speed and clarity over depth. "Clever" lost to "clear" every time we tested. Dense tools need a fast entry point, not a gentle one.
03
Async-first is its own discipline. Three time zones forced every Figma frame to stand on its own — no walkthrough to bail you out.
04
Releases compound differently than sprints. The biggest R2 wins came from watching real friction in R1 and designing against it — not from the next item on the backlog.
Current state · NDA
Specific screens, data flows, and product family details are under NDA.
Case study
Conversational commerce · LATAM
Zubale — Consumer Connect
Lifted user‑attributed sales by up to 18.5% for Latin America's leading retail‑tech platform. The whole store lived inside WhatsApp, where one extra message costs a conversion — so the work was copy, sequencing and CRO, not screens.
+18.5%
user‑attributed sales lift
100+
retailers · 6 LATAM countries
RoleProduct Designer — Consumer Connect
TimelineOct 2023 — May 2025
TeamPM · Engineers · Data · Ops
StatusLive in 6 LATAM countries
The problem
Many retailers, six countries, one chat thread
Zubale needed a unified WhatsApp shopping experience that worked across 100+ potential retailers in 6 LATAM countries — without each retailer building their own conversational flow.
What I designed
Conversational flow patterns, copy systems, and conversion optimization
Redesigned cart, checkout, and post-purchase flows. Authored the UX writing voice + tone for all 6 markets. Iterated on conversion choke-points with A/B tests.
Outcomes
+18.5%
User‑attributed sales lift
100+
Potential retailers
6
LATAM countries
2yr
Tenure on Consumer Connect
Retailers shipped
Walmart
Cashi
Decathlon
Subway
Pizza Hut
Sbarro
Hubii
Angeloni
Condor
Makro
Paris
Farmapronto
In progress
SaaS · Fitness operations
La Central — gym operations in one place
Cut front‑desk check‑in from ~25s to 6s and put every failed payment in a single view. A workspace for independent gym owners, replacing five disconnected tools with memberships, money and the daily floor in one place.
6s
front‑desk check‑in (vs ~25s)
100%
of test users found at‑risk members unaided
RoleProduct Designer · UX Lead
CategorySaaS · Fitness Operations
Year2026
StatusIn progress
RightsProtected — all rights reserved
The problem
Five tools, one frustrated owner
Independent gym owners run their business across five disconnected tools: a POS for payments, a spreadsheet for members, a calendar for classes, a paper clipboard at the door, and a notebook for what didn't work. Revenue leaks through failed payments nobody chases.
Target user
Independent gyms that can't out-spend the chains
Owners and operators of independent gyms, boutique studios, and small multi-site franchises (1–5 locations, 100–800 active members). Software has to earn its keep on day one — no consultant required.
Design approach
Workflow‑first, not feature‑first
01
Research. 12 owner & front‑desk interviews; mapped the real daily workflow against incumbent software.
02
IA reset. From feature‑by‑feature nav to workflow‑first (Today · Members · Schedule · Money).
03
Prototyping. Mid‑fi tests with 5 owners; iterated until time‑on‑task dropped 40%.
04
System. Lightweight design system in Tailwind v4 keeping front‑desk density and owner-dashboard hierarchy consistent.
Outcomes
6s
Front‑desk check‑in time (vs ~25s in incumbents)
1 view
To surface every failed payment
100%
Of test users completed find-at‑risk task unaided
Prototype
Built in Next.js, React, Tailwind, TypeScript
Mocked data · for demo
In production
PropTech · Los Angeles
Conscious Communities
Turned property search and multi‑strategy financial modeling into one workflow, now in production in Los Angeles. GCs, investors and flippers compare 8+ strategies per property without ever leaving the deal.
8+
strategies modeled per property
LA
on‑market + off‑market deal flow
RoleLead Product Designer — end‑to‑end
TimelineOngoing — weekly cadence with founders
TeamFounders (investors+realtors+users) · PM · me · small eng team
StatusIn production · LA
The problem
Million-dollar decisions, run on spreadsheets
Small-time GCs, investors, flippers, and realtors in LA were drowning in property listings — hundreds of options across dozens of possible strategies (renovation, ADU, lot split, tear‑down + new SFR, quadplex). The good deals went in a week, and pricing them required juggling Excel models for every renovation, ADU, SB‑9, and tear‑down scenario.
How I worked
Two meetings a week with the founders — who are also the users
The founders aren't just founders — they're active investors and realtors, the platform's primary power users. I introduced a weekly user‑story template they filled in between meetings (what they tried, where they got stuck, what was missing). They surfaced the problem; I owned the solution.
What I designed
Strategy‑first, not listing‑first
Most real estate tools start with the property and bolt analytics on top. We inverted it: global assumptions drive every search, every property is auto‑modeled across renovation/ADU/SB‑9/tear‑down, sensitivity tables show how YOC and ROI move with cost shifts, and comparison is a first‑class view.
What I learned
The competitor checklist was a trap — the real product was smaller than we thought
01
Feature parity isn't the goal — feature fit is. Early on, the founders wanted to bring over many features from platforms they used. Working through the stories, most were noise.
02
When the user is the founder, separation of concerns becomes your job. The user‑story template kept them as users describing problems, not clients writing specs.
Outcome
B2B
GCs · Investors · Flippers · Realtors
8+
Strategies: Renovation · ADU · SB‑9 · Tear‑down
KPIs
Calculator + sensitivity tables
LA
On‑market + off‑market deal flow
Designed for & with·NVIDIA · Zubale · Codebranch · Pari · Televisa · Walmart
How I work
Strategy, curation, AI, global
01
Strategy
I make the hard calls — what to build, what to cut, why it matters.
02
Curation
Every flow, every word, every interaction — intentional, not accidental.
03
AI-Powered
Claude, Figma AI, prompt engineering — a superpower, not a crutch.
04
Global
US, LATAM, worldwide teams. Multicultural collaboration is second nature.
Experience
Where I've been
May 2025 — Present
Product Designer
Codebranch (US–Colombia)
Design studio with clients including NVIDIA. Leading product design for AI-powered sports apps, a real estate platform, and enterprise tools for NVIDIA Operations.
Claude CodeClaude DesignFigma
Oct 2023 — May 2025
Product Designer — Consumer Connect
Zubale
WhatsApp-based shopping flows for 100+ retailers across 6 LATAM countries. UX writing, CRO, and flow redesign that increased user‑attributed sales by up to 18.5%.
FigmaWhatsApp BusinessMeta
Jun 2023 — Dec 2023
Design Project Manager
Pari Delivery — Mexico
Led the parallel redesign of three product surfaces as Pari pivoted from dark stores to public-facing retail: 15 stores across CDMX, Guadalajara & Querétaro; brand identity; and the customer app shipped to iOS & Android.
ArchitectureBrand DirectionFigma
Previous career
Architect — Retail & Entertainment
Mexico City & Colombia
50+ retail stores across 5 countries. TV sets for Televisa. 15+ public projects in CDMX including PILARES schools.
AutoCADSketchUpRevit
Tools & skills
What I work with
Design & AI
FigmaClaude CodeClaude DesignAdobe Suite
Languages
English (C1, IELTS certified)Spanish (native)Portuguese (basic)Chinese (basic)
Get in touch
Let's build something together
Currently open to remote product design roles. Let's talk about what I can bring to your team.
