Leading engineering at a fintech startup Remote from Argentina (UTC-3)

Gonzalo Martin Perez

I build LLM agents for production, and the evals that show they work.

Explore my work Get in touch

AI Engineer at RampyOpen to remote work

Current role

LLM agents in production at Rampy

Sep 2026 – Present

AI Engineer · Rampy

I lead engineering at Rampy and work closely with the three founders and with specialists in blockchain and integrations. Rampy is an early-stage fintech and blockchain startup whose non-custodial app opens the world of digital assets (swaps, yield vaults, perps and market news) to everyday people through a generative interface that turns complex concepts into simple ones. I own the multilingual AI assistant and the eval suite that gates every release (3,000+ cases, 90%+ pass rate), plus the backend, web and mobile app.

90%+
pass rate on a 3,000+ case eval suite
Global average for production agents on 1,000+ scenarios run in English, Spanish and Portuguese, graded in LangSmith by an LLM-as-a-judge validated against human and product review.
40%
fewer tokens in optimized agent workflows
Measured with load tests and evals after removing redundant tool calls and tightening memory and context; not a benchmark of every query.
2×
faster responses in optimized agent workflows
At least 2×, measured with load tests and evals on comparable optimized agent workflows.
200+
perpetual markets shipped
Hyperliquid markets across crypto, indices, stocks and commodities, alongside swaps across 100+ tokens.

See the Rampy work

Selected work

Case studies and open-source work

Portfolio Assistant

Bilingual RAG assistant with a 1,000-question eval bank

Open-source FastAPI and LangGraph assistant that answers recruiters’ questions about my work, with hybrid lexical and vector retrieval plus GraphRAG (pgvector, Neo4j), cited answers streamed over SSE and per-request cost limits.

Personal open-source project. Implemented and evaluated; not deployed yet.

Example eval case

Question:: Did Gonzalo work at Google?
Expected:: No answer: the public corpus has no source for this employer.
Passes when:: retrieval returns no source at all, so nothing can support a claim.

See the eval bank on GitHub

Source on GitHub

Case study · Client project

Pequeverso: a store with an AI shopping assistant

Pequeverso sells printable learning resources for families with children aged 3 to 7. I built its public store, a private back office and an AI shopping assistant that answers buyers' questions about the catalog.

The assistant: GraphRAG on LangGraph that combines lexical, pgvector and Neo4j retrieval over the store catalog. The server validates every product, price and link before it reaches the buyer, and a monthly spend cap is enforced before each model call.

The assistant is implemented and evaluated; launch pending.

View case study

Featured project

Shipped to production while still at university

Exam administration console with start, finish and defense controls beside live monitor tiles — Managing and monitoring exams

Exam entry card showing the exam code, the student pseudonym assigned for it and the rules — Taking the exam

Correction screen with a student answer beside the yes or no checklist used to score it — Assessing answers

Production product

Jan – Dec 2025

Filomena

Health-sciences exams across five institutions

A legacy CakePHP and jQuery monolith, rebuilt as an API-first platform that now runs high-stakes exams for five Argentine national institutions.

Built by a three-person team. I was the principal author and contributor across architecture, backend, frontend, infrastructure, security, observability and coordination. The platform is a team achievement.

Laravel
React
Next.js
TypeScript
MariaDB
Redis

Read the case study Source on GitHub

5
institutions in production
UNS, UNRN, UNC, UNVM and FAMFyG.
1,000+
simultaneous users
Observed in production with complete data consistency, not a synthetic benchmark.
<300 ms
critical endpoints
Average latency, through SQL tuning, Redis caching and queues.

Experience

Products, teams and contributions

Sep 2026 – Present
AI Engineer · Rampy
Multilingual AI assistant, eval suite and the trading product around it, at an early-stage fintech and blockchain startup.
Dec 2025 – Sep 2026
Software Engineer · Teamcubation
The promotions platform for Payway, a leading Latin American fintech, serving an ecosystem of 350,000+ merchants, 70+ card issuers and 13 countries.
Jan 2024 – Present
AI Engineer & Full-Stack Developer · Self-employed
Occasional, project-based work for e-commerce and digital-product clients, focused on AI automation and shipping complete products quickly.
Dec 2024 – Dec 2025
Software Engineer · Cooperativa Obrera
A permissions-management back office for authorized administrative staff, system administrators and managers at Argentina's largest consumer cooperative: 150+ branches, 2.7M+ members and an ecosystem of 50+ internal systems.

Full experience and projects

Gonzalo Martin Perez, AI Engineer | Software Engineer | Product Engineer

Who I am

From AI agents to the product around them

At Rampy, an early-stage fintech and blockchain startup, I lead engineering. I built a multilingual AI assistant that passes 90%+ of a 3,000+ case eval suite and answers with native UI that links straight into swap, vault and perps flows.

Most of my work is with LangGraph, Claude, RAG and LangSmith: agents that call real tools, retrieval grounded in the business context and evals that decide what ships. I also build the APIs, back offices, web and mobile apps around them, from user needs and architecture through testing, deployment and monitoring.

My open-source Portfolio Assistant shows the same approach in public code: a bilingual RAG assistant with a 1,000-question eval bank.

Currently: AI Engineer at Rampy
Experience: 2.5+ years, since January 2024
Based in: Bahía Blanca, Argentina (UTC-3)
Languages: Spanish (native) · English (B2)

Evals first

Quality is measured on representative scenarios, and evals gate each release.

Grounded answers

Retrieval and tools instead of guesses, with clear limits on what an agent can do.

Latency and cost budgets

Response time and tokens are product metrics, checked with load tests and evals.

Full ownership

From the agent and the API to the app and its behavior in production.

See how I apply them →

From idea to system

Grouped by what it does

Each tool links to the role or project where I used it.

Explore technologies and applications →

Languages & runtimes

Backend & APIs

Frontend & product interfaces

Delivery & collaboration

.tool-carousel_section__09yEx[data-animated="true"] .tool-carousel_viewport__j8ER5 { overflow: visible; mask-image: none; } .tool-carousel_section__09yEx[data-animated="true"] .tool-carousel_rail__veI9P { display: grid; width: auto; } .tool-carousel_section__09yEx[data-animated="true"] .tool-carousel_rail__veI9P li { width: auto; will-change: auto; } .tool-carousel_section__09yEx[data-animated="true"] .tool-carousel_tile__9lLbg { flex-direction: column; align-items: flex-start; } .tool-carousel_section__09yEx[data-animated="true"] .tool-carousel_tile__9lLbg > span:last-child { white-space: normal; } .tool-carousel_control___so1Y { display: none; }

Tools I build with

The full toolkit at a glance. The stack page links each one to where I used it.

Explore the full stack

Education

Information Systems Engineer, UNS

Graduated from Universidad Nacional del Sur in October 2025 with an 8.67/10 average across 34 of 34 required courses, and a final-year project graded 10/10.

8.67 / 10
average grade
Across the full degree, on the Argentine 1–10 scale.
34 / 34
required courses passed
Every course in the program, with no outstanding subjects.
10 / 10
final-year project
Filomena, graded at the highest grade.

Academic record

Let's talk

Building an AI product and need someone who can own it from the agent to production?

Let’s talk about your product.

I’m open to fully remote opportunities.

Email me All contact options