germán tarnoski

AI Evaluator - Software Engineering (Project-Based Contractor)

Alignerr

Jul. 2026 – Sep. 2026 · Remote

Design real-world software engineering tasks from open-source GitHub repos to evaluate frontier coding agents' behavior. Ran each task against different models, graded their behavior against a strict framework (staying in scope, root-cause diagnosis, and other criteria), and wrote technical reports on their performance. Identified patterns regarding where frontier coding agents do well, where they fail, how to optimize token usage, and more.

Freelance Full Stack Developer

MMS Logística

Aug. 2025 – Apr. 2026 · Buenos Aires, Arg.

Designed and built a logistics management platform from scratch as the sole developer, replacing disconnected Excel spreadsheets and end-of-week email updates at a pharmaceutical distribution company. React + TypeScript frontend, Supabase (PostgreSQL) backend. Handles ~400 shipments per week across 11 drivers with real-time multi-user shipment tracking and full lifecycle management, a client file import pipeline with staged review and approval, SAP integration via Supabase Edge Functions triggered by Postgres triggers on every status change (with OAuth2 token caching), PL/pgSQL functions for atomic batch approval and duplicate detection, automated delivery emails via Resend, a JWT-authenticated endpoint for programmatic proof-of-delivery PDF retrieval by external systems, a complete per-shipment audit trail, role-based access separating operator and client views, per-user configurable Excel exports via ExcelJS, and a Playwright E2E suite covering the full lifecycle.

Full Stack Engineer

AlSur

May 2024 – Aug. 2025 · Buenos Aires, Arg.

Joined AlSur, an early-stage startup for travel agencies in Patagonia, shortly after its founding, and led the design, development and deployment of its multi-tenant SaaS platform, replacing WhatsApp threads, paper notes and spreadsheets with a centralized system built with React and Supabase (PostgreSQL). Took technical lead over a second developer, setting direction, reviewing output and mentoring on the stack, while guiding the non-technical co-founder through manual QA and Cursor-assisted prototyping. Built the full reservation lifecycle: a multi-step booking workflow, calendar management for vehicles, guides and equipment, and a complex pricing engine with multi-currency support. Established Row-Level Security for tenant data isolation and a Playwright E2E suite. A simplified version became my university final project, and I graduated with it with excellent feedback; the platform later reached full production-ready status.

Expert AI Data Trainer (Project-Based Contractor)

Invisible Technologies Inc.

Jun. 2023 – Jun. 2026 (intermittent) · Remote

Engaged periodically as an independent contractor, depending on project and work availability, training and evaluating frontier AI models across a wide range of capabilities, including natural conversation, tool use, voice interaction, mathematics and multilingual reasoning (English, Spanish, Programming) through SFT, RLHF, and RAG workflows. Worked across projects with agent pools of up to 500, repeatedly hand-picked alongside 4–5 others for client-requested pilot batches and new capability launches. Served as Quality Analyst in a specialized mathematics project and was consulted by project managers on quality standards and team-level decisions. Guided onboarding of incoming agents across multiple initiatives.

Data Analyst

Coordline S.R.L.

Mar. 2023 – Oct. 2023 · Buenos Aires, Arg.

Automated complex client-data normalization, transformation, and enrichment processes using Python and Pandas, reducing 4–5-hour manual processes to just a few minutes.

Freelance Data Analysis & Automation Support

Decormec S.A.

Aug. 2022 – Mar. 2023 (intermittent) · Buenos Aires, Arg.

Built automation and web-scraping tools (Python, Selenium, Django) for internal data processing. The company's customer order costing web application (labor hours, machine expenses, inventory) was being built by a non-technical team member; I guided that development by teaching Python and Django, establishing a workable structure, and reviewing the work to ensure it held up technically.