$ whoami

Pedro Curti

Software Engineer

Buenos Aires, Argentina

status:online
scroll

// about

$cat about.md

Software Engineering undergraduate at ITBA (thesis pending) with full-stack experience developing production backend features, including batch REST endpoints, event-driven architectures, and hexagonal module design. Skilled in React, TypeScript, Node.js, and PostgreSQL, with a focus on clean architecture and data consistency. I particularly enjoy designing backend solutions, and I’m currently open to new opportunities.

You’ll find me at the gym or rewatching Howl’s Moving Castle for the n-th time (or any Studio Ghibli movie :P)

Languages

TypeScriptJavaScriptPythonJavaC/C++SQL

Frontend

ReactTanStack QueryZustandShadcn UI

Backend

Node.jsExpress.jsZodSpring

Databases & ORM

PostgreSQLMySQLDrizzle ORMHibernateMongoDB

Tools & Practices

GitDockerNeovimCI/CDOpenAPI/Swagger

Methodologies

Scrum

// experience

573eb81(HEAD)Software Developer — Kaizer

Nov 2025 — Jul 2026

Buenos Aires, Argentina

Developed and shipped production backend features for a fitness platform, including batch REST endpoints for exercise operations and soft-deletion flows with dynamic overrides. Architected a transactional abstraction using post-commit hooks to safely decouple side effects from core database writes, ensuring data consistency. Designed an event-driven architecture for cross-cutting concerns such as transactional emails and notifications, and delivered a hexagonal mailing module with localized React Email templates (EN/ES/PT).

TypeScriptNode.jsExpressZodPostgreSQLDrizzle ORMReactTanStack QueryZustandReact EmailResend

// projects

~/projects/rent-price-estimator

Rent Price Estimator

XGBoost regression model on ~48,800 listings to estimate monthly USD rental prices.

Trained an XGBoost regression model on ~48,800 listings across 101 cities to estimate monthly USD reference prices for apartment rentals, achieving a test MAE of ~$163 with a calibrated prediction interval at ~90% coverage.

Ensured robustness through geographic feature engineering (price KNN, city median aggregates, distance to center) fit on training data only to avoid data leakage, hyperparameter tuning via GridSearch, and model deployment behind a minimal HTTP front-end with out-of-scope abstention. The engineered price-KNN feature ranked #1 by permutation importance, raising MAE by ~$266 when removed.

PythonXGBoostscikit-learnpandas
~/projects/discussion-forum

Discussion Forum

Reddit-inspired forum with auth, moderation, and post rating.

Designed a Reddit-inspired forum with user authentication, email notifications, community moderation, role management, and post rating/filtering. Architected a layered RESTful API with clean separation between front-end and back-end.

JavaSpringReactHibernateBootstrap
~/projects/mapreduce-pipeline-with-hazelcast

MapReduce Pipeline with Hazelcast

Distributed MapReduce filters in a cluster environment.

Developed distributed MapReduce filters in a cluster environment with multiple query strategies and concatenated pipelines. Benchmarked and compared strategies, including data upload optimization prior to processing.

JavaHazelcastPython