Industrial engineer, master's in digital transformation and business analytics. I identify the opportunity, quantify the return, deliver the MVP and drive the solution to real adoption — no hype.
8+ years in regulated, data-intensive industries (pharma, FMCG, industrial services). Today I combine an executive role — where I autonomously deliver AI and automation initiatives — with an independent consulting practice, including the design of a corporate AI course for a multinational pharmaceutical company.
No in-house fleet management system; manual processes with no traceability.
Product Owner and developer: SRS, backlog, an 18-entity schema and API contracts, with Claude Code as copilot at every step.
A working application built from scratch — Phase 1 architecture delivered and evolving.
~1,200 pharmacy and administrative staff needed to adopt AI in a regulated sector, with no guide on where to use it — and where not.
I designed the program and content architecture for administrative and operational use cases only, explicitly mapping where AI must not be used: clinical judgement stays with people.
An applied training program for ~1,200 people, with the regulatory boundary defined as part of the deliverable.
~12,036 invoices to migrate manually, with supporting documents scattered across Gmail and no priority logic.
A NestJS + BullMQ architecture integrated via API, plus automated bulk download from Gmail with state-based resumption and hash validation.
Mass migration executed with priority logic, no manual data entry and no dependency on IT.
Manual reconciliation against the carrier's statement: duplicates, seals already invoiced, waybills crossing month-end — a real risk of double billing.
I functionally led an application that downloads waybills via API, cross-checks every seal against the statement and history, classifies each one and generates a consolidated PDF plus the exceptions report.
A faster, traceable month-end close with no double billing and no unsupported charges.
Daily quality-control forms (tanker, tanks and dispensers) were filled in by hand: ~20 minutes a day and transcription errors.
I designed and built end to end a system that generates and files the forms; the signature is only inserted when the form is compliant — judging the fuel stays human.
From ~20 minutes a day to under 5, with no errors recorded since go-live.
Non-standardized contracts and certificates, errors from reusing templates, and a slow process that was hard to audit.
I led a generator built on deterministic merge: it never rewrites clauses, only substitutes fields — every document is reproducible and auditable. Validated with legal counsel.
From hours to minutes per document, zero transcription errors, a complete catalogue ready for signature.
A newly opened business with three lines and no quantitative answers: is it viable? Where is break-even? How long to recover the investment?
As lead consultant I built a 24-month financial model with parameterized scenarios, and a commercial strategy with KPIs tied to the model.
Break-even and verifiable targets from day one — and the methodology became a replicable playbook.
A brand at risk of being dropped by a key client due to falling sales.
Category analysis: I showed the buyer with data that the decline was segment-wide and that my product was cushioning the fall.
Distribution recovered at 100% of points of sale and the portfolio expanded from 1 to 4 SKUs.
Trade marketing teams decide when to activate discounts with no predictive support by category and chain type.
Design, training and validation of an ML model on historical sales and category patterns, after a systematic review of predictive models applied to retail.
Actionable recommendations on the optimal timing of promotional activation to maximize return on investment.
I map the process on the ground and find where AI or analytics changes the outcome — not where it sounds good.
Output · prioritized opportunityA business case with estimated ROI before writing a single line of code. If the numbers don't work, it doesn't happen.
Output · business caseA working solution in weeks: an automated flow, a tool or a product. No decks in between.
Output · solution in productionI train, measure actual usage and adjust until the team uses it without me.
Output · measured adoption