AI Researcher (Audio/Voice)

Location: San Francisco, California, United States

Department: AI Research

Workplace: on_site

Employment Type: full

Description

THE OPPORTUNITY

We are Amplifier, and we have built the world’s first Large Acoustic Model (LAM), a foundation model that uses human voice to detect health conditions. This is sci-fi becoming reality. We have raised significant capital from top-tier investors to turn this technology into a massive new category in healthcare.

We are looking for a researcher who is tired of the "publish or perish" cycle and wants to build intelligence that actually works in the real world.

THE REALITY

Let’s be clear about what we are signing up for.

We are entering a phase of hyper-growth. We are pushing ourselves—and this technology—further than most would consider reasonable. We are doing this because we believe the outcome (saving lives at scale) is worth the intensity required to get there.

• We work in person in San Francisco. We believe that the hardest problems are solved at a whiteboard, not over a Zoom call. We want the energy, the speed, and the camaraderie that comes from being in the arena together.
• We move fast. We don't spend months on theoretical proofs. We hypothesize, we code, we train, and we validate. The feedback loop is immediate.
• We have fun. We are a small, tight-knit crew on an adventure. We work hard because we love the game, not because we have to.

THE MISSION

You will join our elite AI Research team to advance the state-of-the-art in acoustic modeling. You won't just be fine-tuning off-the-shelf models; you will be designing novel architectures that can extract clinical-grade biomarkers from raw audio waveforms.

The Challenge:

• Novel Architectures: Voice is not text. You will push the boundaries of how Transformer architectures process long-range acoustic dependencies and non-verbal signals.
• Biomarker Discovery: You will design experiments to isolate specific acoustic features (jitter, shimmer, respiratory rate) that correlate with health conditions, often discovering signals that medical science hasn't yet quantified.
• Data Efficiency: We are building a foundation model. You will work on self-supervised learning techniques to leverage massive amounts of unlabeled audio data.

Requirements

WHO YOU ARE

• A Builder-Researcher: You have a deep theoretical understanding of Deep Learning (maybe a PhD, maybe not—we care about ability, not pedigree), but you express your ideas in PyTorch, not just LaTeX.
• Audio Native: You understand the physics of sound. You know your way around DSP (Digital Signal Processing), STFTs, Mel-spectrograms, and the unique challenges of modeling raw audio.
• First Principles Thinker: You don't just import HuggingFace libraries. You understand the math behind the attention mechanism and can modify it when standard approaches fail.
• Hungry: You have a chip on your shoulder. You want your work to result in a product used by millions, not just a citation in a journal.

Benefits

WHAT WE OFFER

• Impact: The chance to build a product that literally saves lives.
• Equity: Real ownership. We are early enough that your equity package has life-changing potential.
• The Team: You will work directly with the Founders (Jeremy, Amit, Peh) and our AI research team. No middle management. Just builders.
• Resources: We are well-capitalized (oversized Seed), giving us the compute resources (H100 clusters) we need to execute.

HOW TO APPLY

Don't send a generic cover letter.

Send us your GitHub or a link to a paper/project you implemented. Tell us about a time you had to abandon a standard architectural approach because it didn't work for your specific data problem.

[email protected]

Come build with us.