Staff Software Engineer (AI Evals)

G2i

InternationalÉtats-UnisTemps pleinInformatique et numériqueTélétravailPubliée le

Télétravail ouvert aux candidats de : Cameroun, Afrique du Sud, Algérie, Angola, Bénin, Botswana, Burkina Faso, Burundi, Cap-Vert, Comores, Côte d'Ivoire, Djibouti, Égypte, Érythrée, Eswatini, Éthiopie, Gabon, Gambie, Ghana, Guinée, Guinée Équatoriale, Guinée-Bissau, Kenya, Lesotho, Libéria, Libye, Madagascar, Malawi, Mali, Maroc, Maurice, Mauritanie, Mayotte, Mozambique, Namibie, Niger, Nigeria, Ouganda, République centrafricaine, République démocratique du Congo, République du Congo, Réunion, Île de la, Rwanda, Sahara occidental, Sainte-Hélène, Ascension et Tristan da Cunha, Sao Tomé-et-Principe, Sénégal, Seychelles, Sierra Leone, Somalie, Soudan, Soudan du Sud, Tanzanie, République unie de, Tchad, Togo, Tunisie, Zambie, Zimbabwe, Åland, Îles, Albanie, Allemagne, Andorre, Autriche, Bélarus, Belgique, Bosnie-Herzégovine, Bulgarie, Chypre, Croatie, Danemark, Espagne, Estonie, Finlande, France, Gibraltar, Grèce, Guernesey, Hongrie, Île de Man, Îles Féroé, Irlande, Islande, Italie, Jersey, Lettonie, Liechtenstein, Lituanie, Luxembourg, Macédoine du Nord, Malte, Moldova, République de, Monaco, Monténégro, Norvège, Pays-Bas, Pologne, Portugal, Roumanie, Royaume-Uni, Russie, Fédération de, Saint-Marin, Saint-Siège (état de la cité du Vatican), Serbie, Slovaquie, Slovénie, Suède, Suisse, Svalbard et île Jan Mayen, Tchéquie, Ukraine

Description du poste

Evaluate the quality of interactions with modern coding agents such as OpenAI Codex and Claude Code. This is not a traditional engineering role: you will assess whether models think and communicate like great engineers, rather than write production code. You will evaluate AI-generated coding interactions in real-world scenarios, focusing on whether responses are useful, correct at a high level, and aligned with strong engineering judgment. You will assess explanations and reasoning, distinguish levels of response quality, provide direct feedback, and help define what good interactions with tools like Cursor look like. The work requires making subjective but rigorous judgments about whether explanations are helpful, whether models guide users well, and whether interactions build or erode trust. Engagement details: Contract; approximately 20 hours or more per week; through October, with possible extensions; start as soon as possible. The selection process includes a take-home evaluation exercise, a background check, and an offer.

Profil recherché

Staff / Principal-level engineer or equivalent experience. Strong background in Go, Python, Rust, TypeScript, or JavaScript. Hands-on experience using OpenAI Codex, Claude Code, and Cursor. Deep familiarity with modern AI-assisted development workflows. Able to evaluate code without needing to execute it or review every line in depth. Comfortable giving direct, opinionated feedback and maintaining a high bar for engineering quality. Nice to have: Experience with Cursor or similar AI-first IDEs; prior exposure to prompt design or evaluation workflows; experience mentoring senior engineers or defining engineering standards.

Offres similaires

Toutes les offres — Informatique et numérique