Research Lead – Pre-training Safety

🔥 il y a 18 heures

🌐 États-Unis, Singapour – Distant

🇺🇸 États-Unis – Télétravail 🇸🇬 Singapour – Télétravail

🏄 California – Distant

🏄 California – Remote

💵 $290 000 - $450 000 / an

⏰ Temps Plein

🟠 Senior

👻 Score fantôme 10%

Ce que ça veut dire : la probabilité que cette offre reste ouverte sans recrutement actif.Comment le lire : 60 % ou plus, on pense que c'est une offre fantôme. Un score plus bas est moins inquiétant.Comment on la calcule : ancienneté, republications, postes ouverts vs taille de l'équipe, rapidité de clôture, et si le salaire est indiqué.À noter : une estimation à partir de données publiques, pas un jugement sur l'entreprise. De vraies offres peuvent avoir un score élevé, alors prenez-le comme un indice, pas comme un fait.

Postuler Maintenant

Trouver des Emplois à Distance Similaires

Configurer une alerte email pour les nouveaux emplois

📊 Vérifiez votre score de CV pour ce poste

Améliorez vos chances d'obtenir un entretien en vérifiant votre score de CV avant de postuler.

Télécharger le CV

FAR.AI

Site WebLinkedInTous les Emplois

11 - 50 employés

Fondée en 2022

🤖 Intelligence artificielle

📚 Éducation

🤝 À but non lucratif

Artificial Intelligence • Education • Non-profit

Ashby Electrical Limited est une entreprise d’électricité fondée en 2018, spécialisée dans la fourniture de solutions électriques économiques et durables pour des projets résidentiels et tertiaires. L’entreprise se distingue par la qualité de ses prestations tout au long du cycle de vie des projets, incluant la conception, l’installation, les essais et la mise en service. Ashby Electrical s’engage à bâtir des relations solides avec ses clients, fondées sur la confiance, l’intégrité et le professionnalisme, afin de livrer des projets dans les délais et dans le respect des budgets.

Description

• Develop and lead FAR.AI’s pre-training safety research workstream • Direct capability-control research focused on removing harmful capabilities while preserving benign ones • Scale methods such as Deep Ignorance to models with over 100B parameters and over 1T tokens • Partner with the red team to stress-test resulting models and analyze scaling to frontier systems • Research improved data filtering using data attribution, influence-based selection, and sophisticated classifiers • Use gradient routing to isolate dual-use capabilities in model components such as MoE experts • Develop training methods to actively remove harmful capabilities through unlearning and next-token prediction • Add synthetic data to pre-training or mid-training to shape model representations and behavior • Build and lead the team and set its research direction • Mentor Members of Technical Staff and remain hands-on by writing code and running experiments • Articulate a research agenda and theory of change for mitigating catastrophic AI risks or increasing beneficial outcomes • Lead novel research projects with potentially unclear progress or success markers • Share findings through academic publications, blog posts, ML conferences, and policymaker briefings • Contribute to FAR.AI’s intellectual environment and research culture • Build a research field through grantmaking and events • Connect research to real-world deployments through independent testing and government advising

🎯 Exigences

• Strong existing research track record in AI or another highly technical subject, such as computer science, mathematics, or physics • Deep experience with language-model pretraining, dataset construction, or controlled training experiments • Experience building large-scale pipelines for scoring, filtering, deduplicating, and sampling training corpora • Strong experimental judgment, including safety–capability evaluations, distribution-shift analysis, and statistically rigorous model comparisons • Ability to build and debug research systems directly, from classifier fine-tuning through distributed training and evaluation • Either a clear research agenda with a theory of change, or a strong track record and research space to develop into an agenda • Experience leading a team, mentoring graduate students, or supporting early-career researchers; informal leadership counts • Ability to communicate novel methods and solutions to technical and non-technical audiences • Substantive engagement with machine learning research through prior research, employment, or sustained independent contribution; not a new entrant • Established AI safety publication record preferred • Comfort writing grant proposals and navigating external collaborations preferred • Full-time availability, 40 hours/week • Ability to work remotely in most countries or in person in Berkeley, California or Singapore

🏖️ Avantages

• Competitive salary of $290,000–$450,000/year depending on experience and location; exceptional candidates may receive a higher salary range • Work-related travel and equipment expenses covered • Catered lunch and dinner at Berkeley offices • Flexible remote or in-person work options • Visa sponsorship for in-person employees • Paid 3–5-day work trial • Sizable compute budgets on a managed cluster • Dedicated engineering team supporting compute infrastructure and experiment scaling • Opportunities to collaborate with governments, leading AI companies, and academics • Professional development through research leadership, mentoring, publications, conferences, grantmaking, and events