AI Researcher (On-Device AI)

Location: San Francisco, CA, USA; in person

Company

hirify.global is a stealth AI company building trustworthy, consumer-grade agents that bring frontier-model capabilities to phones, laptops, and wearables.

What you will do

• Lead one or more model capabilities end-to-end, from data mixtures and training objectives through evaluation and production deployment.
• Own training workstreams with clear success metrics and ship checkpoints into on-device runtimes.
• Design experiments and writeups that identify high-leverage research directions and improve team decision-making.
• Support infrastructure and product engineers in turning research results into shipped capabilities.
• Advise partnerships on feasible capabilities for upcoming device refreshes.
• Review and improve research code across the team.

Requirements

• Experience with model training techniques including distillation from frontier teachers, small-scale MoE, speculative decoding, and KV-cache compression.
• Ability to design experiments around meaningful evaluation metrics.
• Experience or interest in pretraining, SFT, RLHF, DPO, GRPO, distillation, and quantization.
• Ability to work in person in San Francisco.
• Ability to take models from training runs through evaluation and on-device deployment.

Nice to have

• Experience with on-device tool use and agentic post-training.
• Experience deploying production models under hardware deadlines.
• Experience working across the full stack from training runs to phones.

Culture & Benefits

• Work on consumer-scale AI agents designed around trustworthiness and reliability.
• Competitive cash compensation and meaningful equity.
• Top-tier relocation and immigration support.
• Permission to publish research that is safe to publish.
• Work with researchers and engineers from Stanford, OpenAI, and DeepMind backgrounds.

Hiring process

• Submit a link to an interesting result, such as a paper, blog post, model card, or GitHub project, with a paragraph explaining its importance.
• Include a resume, Google Scholar profile, or LinkedIn profile.
• Exceptional candidates receive a response within 48 hours.