AI Researcher (RL + On-Device)

AGI, Inc. is a SF neolab building trustworthy, consumer-grade AI agents that run on-device. Founder Div Garg pioneered browser-use agents, he built the first agent in Silicon Valley to pass the California DMV's online test, and raised a $30M Series A (General Catalyst, Khosla) at his prior startup before spinning out AGI, Inc. as a research lab. AGI has held #1 on the AndroidWorld leaderboard, the industry standard for mobile-use performance, since October 2025, is backed by Menlo Ventures, Point72 Ventures, and the Anthropic Anthology Fund (plus senior leaders from DeepMind and OpenAI), and just added the co-founder of Android as an advisor.

Headcount has more than doubled in six months, with a seed round closing by year-end. AGI builds small action models and an execution layer that let agents see the screen, understand context, and act across any app, no APIs, no per-app integrations, running privately, on-device. Uniquely, partners come to them: AGI works with Qualcomm to optimize and deploy its agents on Snapdragon-powered edge devices (phones, laptops, smart glasses, watches) and with Lenovo (a Lenovo Qira proof-of-concept demoed live at MWC 2026), among other large partnerships that are official but not yet public.

The team runs two lanes, a B2C lane (iOS/Android apps) and a B2B lane (device-maker partnerships), out of an on-site SF office. The bar: everyday AGI in a hundred million pockets.

This is the researcher who's tired of training models that go behind an API and wants their model on the device in a hundred million pockets. The problem: frontier capability inside the compute and memory envelope of a consumer device, phone, laptop, wearable, which AGI considers the most interesting research problem in applied AI today. You'll lead training for one of the model families that powers AGI's on-device agents: pretraining recipe choices, post-training (SFT, RLHF, DPO, GRPO, and whatever's next), distillation, quantization, and the long tail of tricks that make a small model punch above its weight.

Pedigree (which school) matters least. The archetype is someone who's seen fast scale at a frontier lab (ex-OpenAI, ex-Perplexity) and/or ran their own startup or was an early hire somewhere that scaled, and, as everywhere at AGI, brings real agency and urgency.