SoftwareEngineer

The problem

Small and mid-size Argentine farms plan the year on spreadsheets and experience. There's no accessible way to ask "what if" — a dry season, a commodity-price crash, switching a field from crops to grazing — before committing real money to the decision.

My solution

• →Multi-tenant SaaS where each producer models their own operation: fields, herds, plantings and cash flow.
• →A what-if simulation engine covering five scenarios: yield variation, commodity prices, crop rotation, cattle fattening and adverse climate.
• →A machine-learning service that predicts departmental crop yield from public MAGyP and weather data.
• →An LLM step that turns each simulation into a written, plain-language advisory report.

Architecture

1. 01Front end — React 19 + TypeScript + Tailwind, deployed on Vercel
2. 02API — Java 21 + Spring Boot 4 REST service with JWT auth, on Render
3. 03Data — PostgreSQL on Neon; Docker Compose for local parity
4. 04ML — Python microservice (FastAPI + scikit-learn); yield model validated with a temporal split

Results

• ✓All five simulation scenarios implemented; full stack deployed on Vercel + Render + Neon.
• ✓Yield model trained on real public datasets and validated with a temporal split, so it's never scored on data from its own future.
• ✓In progress — final-year Computer Engineering thesis, UADE (expected 2027).

The problem

Geoscientists run heavy visualisation software on a shared remote-graphics platform (RGFX) — blade servers, virtual machines and a Cisco cluster. Keeping it healthy and handling every access and setup request by hand is slow and easy to get wrong.

My solution

• →Day-to-day administration of the RGFX platform: servers, virtual machines, the Cisco cluster and the network behind them.
• →Automation in Ansible / Ansible Tower and PowerShell for provisioning, health checks and Windows Server upkeep.
• →Incident and service-request queue handled in ServiceNow, with runbooks for the tasks that keep coming back.

Architecture

1. 01Request — access or a new workstation comes in through ServiceNow
2. 02Provision — Ansible Tower / PowerShell runs the setup and configuration steps
3. 03Verify — health checks confirm the VM and platform services are up
4. 04Document — each manual fix becomes a runbook, then a candidate for automation

Results

• ✓Recurring provisioning and health-check work moved from manual steps to repeatable playbooks.
• ✓Runbooks turned on-call knowledge into something a teammate can pick up.

The problem

The firm had no web presence: no way for prospective clients to find it, read its practice areas or get in touch.

My solution

• →Full site in React, organised around practice areas — criminal, civil, commercial — with clear contact paths.
• →Static generation for fast loads; semantic HTML and metadata for search visibility from day one.
• →Deployed on Vercel with zero-config CI/CD.

Architecture

1. 01Next.js App Router — static generation for instant loads
2. 02TailwindCSS — small, consistent design system
3. 03Vercel CDN — global edge delivery, deploy on push

Results

• ✓Live in production as the firm's main point of contact online.

The problem

The firm's litigation-training course needed its own identity — separate from the law firm's corporate site — that reads as academic and premium, explains the curriculum, and converts visitors into applicants.

My solution

• →An editorial, courtroom-toned landing page — serif type over a real courtroom photograph — deliberately distinct from the firm's own site.
• →A chronological curriculum section: each module as a numbered card on a vertical timeline.
• →A custom interactive diagram — "Anatomía del Juicio Oral" — mapping where the judge, witness, defense and prosecution sit, with a scrollable phase selector.
• →A dedicated admissions form collecting profession, city and motivation, with optional CV upload.

Architecture

1. 01Next.js App Router — same stack as the firm's main site, static generation
2. 02TailwindCSS — a warmer, editorial variant of the shared design system
3. 03Custom SVG/CSS component for the courtroom-anatomy diagram
4. 04Vercel — deploy on push

Results

• ✓Live in production as the course's admissions funnel.