Santiago Varacca
Resume

Santiago Varacca — Software Engineer

I build across product, systems, and AI-native workflows.

Nearly five years shipping B2B, commerce, and operational software across frontend and backend. My deepest production experience is TypeScript full-stack; my recent independent systems are delivered through coding agents, with authorship and evidence made explicit.

Download resumeSee selected workGitHub

Selected evidence

The work that actually defines my profile.

Almost five years across real products, plus a deliberate bet on verifiable agent infrastructure. Smaller tools and unfinished systems stay visible without being inflated into flagship work.

// 01

B2B · commerce · logistics

Production systems across three domains

Professional work · 2021—now

A continuous full-stack track across Wellio, Oh! Gift Card, and Urbano: product interfaces, services, data, permissions, integrations, and operational workflows.

  • React and TypeScript
  • Node.js and NestJS
  • PostgreSQL, SQL Server, MongoDB
  • REST and gRPC
  • Microfrontends and microservices
See experienceDiscuss the work
Why it mattered
These systems support people operations, digital gift-card commerce, and logistics workflows where state, money, and partial failures matter.
My scope
At Wellio I worked across most product modules; at Oh! Gift Card I contributed substantially to frontend and backend; at Urbano I have worked across the main React application and five NestJS services.
Evidence
The Urbano record includes 69 assigned tickets, 357 commits, and work in six repositories. It includes credit/debit-note flows, batch processing, decimal precision, concurrency, and recovery work.
Boundary
The source is private and business outcomes were not measured consistently. Postal Plus was delivered through development/QA and is not claimed as confirmed production impact.
// 02

Governed infrastructure for agent work

Agent Work — Standard and Next

Standard release in preparation · Next in development

A product line for making work across coding agents bounded, attributable, independently verifiable, recoverable, and governable across tools and teams.

  • Task packets, receipts, and handoffs
  • Roles, scopes, and governance
  • Independent verification
  • Recovery and continuity
  • TypeScript CLI, MCP, and runtimes
Ask for a walkthroughAuthorship model
Why it mattered
Agent work becomes unreliable when assignment, evidence, verification, delivery, and takeover remain implicit or live only in a chat transcript.
My scope
Coding agents implemented both repositories. Standard focuses on task packets, receipts, handoffs, verification, and recovery. Next explores organizations, roles, conversations, context, governed work, runtimes, and projections.
Evidence
Standard has executable workflow tooling and a passing automated suite. Next contains the broader system and hundreds of tests, but still needs a narrow end-to-end tracer before it can be presented as a product.
Boundary
This is the deliberate architecture and AI bet. It still has no external adoption, and it does not prove that I manually authored the TypeScript implementation.

Smaller tools and work in progress

These add context to how I explore products and systems. They are not presented as achievements on the same level as my professional trajectory or Agent Work.

  • Rutero

    A city-data and outing-planning system with provenance. It still needs real user validation.

    In progress

  • Aurora

    A governed assistant system centered on permissions, memory, explicit state, and auditable actions.

    Paused · architecture case

  • Changuito

    A small supermarket price-comparison tool implemented by agents for my own use. Useful, but not a flagship project.

    One-day personal tool · live

Experience

Nearly five years across product surfaces and system boundaries.

The recent work is the easiest to quantify, not the whole identity. The through-line is operating across frontend, backend, data, and business workflows.

Buenos Aires, Argentina · English C1 · Computer Engineering at UNLP

  1. 2026 — present

    Urbano Express

    Full-stack Software Engineer

    Work on SUGU, an internal logistics platform, across its main React application and several NestJS services.

    • 69 assigned tickets and 357 commits across six repositories
    • Credit/debit-note workflows and batch billing
    • Async processing, idempotency, decimal precision, and failure recovery

  2. 2023 — 2026

    FK Tech Solutions Factory

    Full-stack Software Engineer

    Product work across Wellio, a B2B people platform, and Oh! Gift Card, a multi-brand digital gift-card commerce product.

    • Frontend and backend work across most Wellio modules
    • Substantial contribution to Oh! Gift Card
    • Microfrontends, microservices, REST, gRPC, PostgreSQL, and MongoDB

  3. 2023

    SwagCharm

    Frontend Developer · remote US team

    Remote product delivery using React, TypeScript, and REST APIs.

    • Cross-border team collaboration
    • Production frontend work and review cycles

  4. 2021 — 2023

    42i

    Frontend Developer

    First professional product experience with React, Redux, APIs, CSS, and Sass.

    • Customer-facing frontend delivery
    • Foundation in component-driven web development

Wellio and Oh! Gift Card are deliberately described at product level until individual decisions and outcomes can be documented without inventing detail.

How I work with coding agents

AI changes how the work gets done. It does not erase responsibility.

My recent independent repositories were implemented entirely by agents. I do not present that code as manually authored. I present the problem, direction, delivered system, evidence, and limits as separate things.

The rule

If I cannot explain or defend a decision without another agent, it is not used as evidence of my technical ownership.
  1. 01

    I define the target

    The useful problem, constraints, risks, acceptance criteria, and stopping point are the human responsibility.

  2. 02

    Agents execute the implementation

    Research, code, tests, and documentation can be delegated. The case says that plainly instead of hiding behind “I built”.

  3. 03

    Evidence stays separate from authorship

    A live system, a passing suite, or a benchmark proves a result. It does not prove who understood every decision or wrote the code.

  4. 04

    Limits remain visible

    No invented users, revenue, production impact, language fluency, or ownership. Unverified claims stay out.

Technical foundation

A deep base, a wider operating range.

I do not claim every language used by an agent as a personal skill. The first two columns are grounded in professional work; the third describes how I now deliver independent systems.

Production depth

  • TypeScript and JavaScript
  • React and Redux Toolkit
  • Node.js and NestJS
  • PostgreSQL, SQL Server, MongoDB

Systems work

  • REST and gRPC
  • Microservices and microfrontends
  • Async jobs and idempotency
  • Permissions, reconciliation, recovery

AI-first delivery

  • Problem framing and constraints
  • Bounded agent workflows
  • Acceptance and evidence
  • Explicit authorship and limitations

Remote roles first · selected client work too

Need someone who can move across product and systems?

I am looking for strong international full-stack or product-engineering roles. I also take selected client work involving operational workflows, data, integrations, or agentic systems.

Email meDownload resumeLinkedIn