- Mexico City
- 10+ years at growth-stage startups
- Operator-founder
- Open to travel: SF, NYC
Solutions Architecture leader for growth-stage and enterprise teams.A decade with customers. Product-minded, AI-enabled leader.
I’m Shirley. I built and led Solutions Architecture, Customer Success Engineering, and Support at Clearbit (acq. HubSpot), and led Solutions Engineering at Demandbase, for tier-one customers like Stripe, Slack, Meta, Intercom, and Segment. Today: five live businesses in Mexico City, and the AI-native systems behind them, built with Claude.
- 10+years in growth-stage and enterprise customer engineering
- 3customer-engineering functions built from zero
- 1,000+customers supported, SMB to enterprise
- 4+production AI systems shipped with Claude
- 5physical businesses created in Mexico City
- 4.8★Google rating across those physical businesses, from 3,000+ reviews
Joined Clearbit pre-Series A and built Solutions Architecture, Customer Success Engineering, and Support Engineering as net-new functions, leading them through the company’s growth phase and supporting tier-one accounts including Stripe, Slack, Meta, Intercom, and Segment. (Acquired by HubSpot in 2023.) Demandbase before that, leading SE for Fortune 500 customers across financial services, technology, healthcare, and retail. Recently shipped four production systems with Claude. The combination I’d bring on day one: a decade building and leading pre-sales customer engineering, plus hands-on Claude experience.
The work,and the proof.
Four things a Solutions Architecture leader owns, and where I’ve already done each one.
- 01 · Build and lead the team
Hire, develop, and guide pre-sales Solutions Architects, with real technical depth and real career paths.
My evidenceBuilding and leading customer engineering teams from the ground up has been the most rewarding part of my career. At Clearbit, I established Solutions Architecture, Customer Success Engineering, and Support Engineering as net-new functions spanning four product surfaces: Salesforce package, APIs, native integrations, and the platform itself. I hired and developed engineers and architects, raised the bar for technical guidance, and created career paths that supported growth from Support roles through Customer Success Engineering and Engineering project leaders. Years of running that through Series A growth is the closest analog to building a team like this one.
- 02 · Sponsor strategic customers
Be the technical sponsor for high-value accounts, from use-case scoping and POC execution through to production.
My evidenceTechnical sponsorship for strategic customers has been the work: owning architecture conversations from initial scoping through POC through deployment, and earning the credibility that drove renewal and expansion at Clearbit and Demandbase. The recent Claude work adds practitioner grounding to those conversations: I’ve shipped enough to speak from experience, not theory.
- 03 · Partner with sales leadership
Find the strategies, verticals, and horizontal use cases that drive adoption, and build the technical case that moves deals to close.
My evidencePartnership with sales leadership has been daily work for a decade. At Demandbase I led pre-sales for ABM solutions across financial services, technology, healthcare, and retail customers, the vertical mix most enterprise teams serve. At Clearbit, owning both pre-sales and post-sales technical work meant operating in step with sales leadership on every strategic deal: sourcing patterns, prioritizing target accounts, building the technical case that moved deals from evaluation to close. The reps trusted me in the room because what my teams and I committed to in pre-sales shipped without surprises post-close.
- 04 · Close the loop with product
Work with product and engineering so customer feedback gets prioritized and blockers to adoption get resolved.
My evidenceClosing the customer-to-product feedback loop has been the core of every customer engineering function I’ve built. At Clearbit I stood up Solutions Architecture, Customer Success Engineering, and Support with explicit feedback pathways, so success metrics, customer development, and escalation patterns all fed product strategy. Across years of running that function, the patterns we surfaced from 1,000+ customer engagements drove what shipped next.
Ten plus years building, integrating, and deploying the tools sales, marketing, and revenue teams rely on.Before AI.
Click any row to expand
Three years outside a tech title.Not one away from the work.
The honest version
In 2023 I stepped away from tech roles, not from the work. The businesses now have a team running them day to day, and I’m ready to lead customer engineering again.
- Since 2023
- My partner and I built five businesses in Mexico City from zero, and I built RIOS, the platform that runs them. Then I engineered myself out of the day-to-day: hired and developed the team, wrote the processes, reduced a majority of the single points of failure. My partner leads the operation now, and the businesses run whether I’m in the room or not.
- Why the skills stayed sharp
- Wiring POS, accounting, payroll, and banking together is the same integration and data work I did for customers like Stripe and Slack, with my own P&L on the line. I kept shipping the whole time: four production systems built with Claude. Being the customer taught me what breaks in production and what actually gets adopted.
- What I’m looking for
- Leading pre-sales Solutions Architecture and post-sales Customer Success Engineering for growth-stage and enterprise customers: the technical conversations, the advisory work, and the path from evaluation to production. Ideally close to AI, helping customers put it to work.
After a decade in tech, my partner and I moved to CDMX and built these from the first hire. The work changed less than you’d think: customers, staff, regulators, and suppliers became the inputs, and the systems thinking stayed the same.
What I built along the way.Four builds in production.
Each one shipped with Claude, live today, and solving a problem I hit first-hand running businesses in Mexico. Pick one to see the architecture, where AI sits, and what it changed.
A complete operating system for food businesses, built with an operator's instincts and a product manager's discipline
Sales, purchasing, invoices, bank, payroll, scheduling, inventory, recipes, and P&L in one place. Underneath, RIOS collects and cross-checks data points from every source to detect where money is quietly leaking, and a completeness layer refuses to show a number it can't stand behind. It started as the internal system for our five businesses and now runs as a public product with a free plan and a paid savings flywheel.
Each business in the portfolio ran on its own mix of POS, inventory, accounting, and staffing tools. Reports disagreed by 5–15% on any given day, and losses hid in the gaps between them: a short cash drawer here, a supplier price creep there. The decisions that mattered (staffing tomorrow, ordering next week, whether last week’s price change worked) were being made on lagged, incomplete, or inconsistent data.
- 01Ingestion
Adapters for POS, CFDI invoices, bank statements, and PDF/CSV imports backfill history on a rolling cursor and catch late-arriving rows. Every run is logged and idempotent: running the same job twice never duplicates data.
- 02Completeness
A completeness layer per company, brand, and location. Every view declares the data it depends on and refuses to render a misleading number when something is missing or stale.
- 03Reconciliation
POS sales, terminal payouts, CFDI invoices, and bank deposits compared in the same units, so the numbers actually line up. When they don't, the gap gets flagged instead of averaged away.
- 04Lossdetection
Independent signals run across POS tickets, voids, discounts, cash counts, invoice lines, recipes, and payroll: short cash drawers, discounts concentrated on one cashier, voided orders that were likely already made, paying a supplier more than your own best price, and ingredient increases eroding plate margin. Every loss is measured in pesos with the receipts attached, opens a tracked episode with an owner, and recovery is re-measured the following month.
- 05Savingsflywheel
Signal, owned action, pesos recovered and verified against real invoices and bank movements. Purchase lines from every business on RIOS also feed anonymized supplier-price benchmarks, shown only once at least five distinct buyers contribute. Each operator who joins makes the alerts sharper for everyone: a network effect on top of the loop.
- 06Executivelayer
A small, opinionated set of views (daily sales against target, ticket size, item mix, labor against forecast, the losses open this fortnight) built around what an operator actually decides on, not what's easy to chart.
- 07Agentlayer
Claude on top, scoped to an explicit memory of business structure, vendors, and known anomalies. Used for explanation, summarization, and drafting weekly operating notes that are reviewed before they leave the system.
Used. Claude handles the unstructured inputs that feed the database (receipts and statements), extracting structured fields so the deterministic side can reconcile against them. Every output is validated before it lands.
- Month-end close happens continuously instead of after a three-week wait.
- Waste down 5% in the first month, driven by real-time margin and cost alerting.
- Losses caught from multiple independent data points (cash, discounts, voids, supplier prices, recipe margins), each with receipts and a measured recovery.
- AI extracts invoices (CFDIs, PDFs, XMLs) and matches them to bank deposits and POS sales. Manual reconciliation is gone: for us, two full-time jobs.
- Labor cost as a percent of revenue tracked continuously per location, not at the end of the period.
- Multi-entity, multi-brand, multi-location P&L in one engine. A new business is online in days.
- Productized at holarios.com: a free plan with no credit card, and a paid savings flywheel that compounds as more operators join.