Isabel Oliveira
SQL · BigQuery · Data Pipelines · Data Modeling · Revenue Data Platforms
I build the data platforms that revenue runs on, and I understand the business behind the numbers I model. Four years took me from AI training to revenue operations to data engineering, which is why I can sit between a warehouse migration and a CEO conversation without changing languages.
São Paulo, Brazil · Open to roles anywhere: remote, hybrid or on-site, relocation included.
ARR, MRR bridge, churn and cohorts are not dashboards to me; they are contracts. I model them so a CFO and a data engineer can both defend the same number.
I report directly to the CEO, turn ambiguous business questions into scoped data work, and take the delivery back up to executive decisions.
I lead a team of 2–3 data engineers and run cross-functional squad initiatives across Engineering, Marketing, Sales and Customer Success.
- Lead the data team (2–3 engineers) and report directly to the CEO, translating between the engineering layer and executive decision-making, and running cross-functional squad initiatives across Engineering, Marketing, Sales and Customer Success.
- Own the revenue data platform end to end: BigQuery as the analytical source of truth, PostgreSQL/Supabase for operational workloads, architected around explicit system boundaries, field-level ownership contracts, reverse-ETL workflows and a unified customer data layer over Guru, Pagar.me, RD Station, Calendly, Typeform, Notion, Meta Ads and Google Ads.
- Built the data warehouse from the ground up on an ELT approach (raw data lands first, transformation happens in the warehouse), with staging and fact layers and pipelines for ingestion, reconciliation, deduplication, qualification and assignment.
- Run an ingestion engine processing roughly 2,000 leads a month, made concurrency-safe with a Supabase-backed mutex (retries and timeouts), under freshness, volume, DLQ and Slack monitoring across 59 n8n workflows and reconciliation jobs.
- Established the company's first auditable ARR, MRR bridge, churn and cohort reporting in BigQuery. Revenue-band coverage went from 47% to 81%; ad-level attribution coverage was validated at 95.8% across 20,709 leads.
- Hardened production through CI policy gates, row-level security requirements, migration controls and peer review. Those processes caught both a cross-account PII exposure and an unsafe production DELETE before either reached customers.
- Built the data and retrieval foundation of an internal revenue intelligence application in Node.js, React and TypeScript, turning recorded sales conversations and customer testimonials into structured, searchable information.
- Designed the processing pipeline end to end, covering call ingestion, transcription, AI classification, topic extraction, objection categorization and product-theme organization, and extended the same architecture to WhatsApp, covering both history and live interactions.
- Delivered an objection-handling system grounded in evidence from real closed deals rather than generic scripts, letting reps retrieve answers by topic, objection and product context.
- Managed revenue operations and an engineering squad, owning capacity planning, resource allocation and sprint execution, and set the governance and prioritization frameworks for ambiguous, high-complexity demands.
- Deployed AI-driven Customer Success support in Botpress that absorbed up to 40% of incoming demand, escalating to human agents on the rest.
- Delivered a MongoDB reconciliation layer syncing lead engagement state between Sales and Marketing: it flagged in ActiveCampaign and later Insider whenever a lead was already in active outreach, suppressing the nurture that used to reach them mid-conversation.
- Directed a lead engagement system in Postgres and Clojure wired into Zenvia Conversion and Pipedrive, and built a delivery performance dashboard in SQL and Looker Studio that made squad throughput measurable for the first time.
- Led the technical preparation to migrate the company stack to HubSpot, scoping data models and integration requirements; investigated production errors in Datadog with SQL root-cause analysis.
- Ran revenue operations across Marketing, Sales and Customer Success, mapping end-to-end workflows and data flows and standardizing them into documented processes, with the backlog triaged by ICE scoring and criticality.
- Designed and maintained the integrations connecting the CRM, messaging platforms and internal systems, and benchmarked vendors (Intercom, Zenvia, Twilio, 360dialog) into platform decisions.
- Automated the creation of thousands of tracked links used by Marketing and Sales, removing manual setup from every campaign launch.
- Consolidated four separate sales tools into a single operations hub, cutting a third-party vendor from the stack and saving roughly $110K in annual licensing.
- Designed the automation architecture behind the sales process, mapping business rules, required fields and branching logic for every edge case before implementation, then shipped it, replacing tribal knowledge with one source of truth for every routing decision.
- Rebuilt the sales chatbot on Zenvia Conversion for stronger native integrations and runtime performance, unlocking routing rules and data capture the previous vendor could not support.
- Cut link generation from 24 hours to 3 seconds by redesigning the process instead of automating the old one.
- Owned sales operations automation end to end in Zapier and custom JavaScript, and integrated the CRM (Pipedrive) with internal tools alongside engineering.
- Built the reporting layer that gave leadership its first real visibility into sales operations.
- Analyzed production conversation data used to train IBM Watson conversational models, owning the analyze → retrain → validate cycle and turning conversation analytics into changes to intents, responses and model behavior.
- Rebuilt the AI model of an at-risk account from insights extracted out of its conversation data; the account became a LATAM success case.
Twitch TV (freelance, 2021–2022) and Liga GG (2021). Where the data started: channel analytics drove what to produce and when to schedule it, and the first contact with programming and JavaScript came from building chat bots on StreamElements and NightBot.
Built from zero at Piece: staging and fact layers in BigQuery, ingestion and reconciliation pipelines, and the company's first auditable ARR, MRR bridge, churn and cohort reporting. Revenue-band coverage 47% → 81%; 95.8% ad-level attribution validated across 20,709 leads.
Under NDA, so no names and no client figures. A production Postgres was serving transactional and analytical load at the same time, and the analytical side eventually degraded the database customers depend on. I designed the separation: a dedicated analytics database fed by selective logical replication, aggregates computed in-database (plpgsql + pg_cron, incremental by watermark), and one versioned SQL view per metric as the semantic contract, so that every metric has a single reviewable definition and ad-hoc analysis never touches production. Scope, architecture decisions and results can be discussed to the extent the NDA allows.
Own product: website + mobile app (PWA) with a custom design system. Native version under review on Google Play.
Website and Instagram automation platform (ManyChat-style), with an operations panel.
Hiring for a data role?
Open to roles anywhere: remote, hybrid or on-site, relocation included.
isaolivcld@gmail.com