Machine Learning Engineer; Voice

Machine Learning Engineer (Voice) - London An AI-native engineering org. Agents across coding, reviewing, QA, ops, support, scheduling. You'll help turn the company itself into an intelligent system. Speech-to-speech agents that feel like a person in the room. Interruptions, back-channels, noisy kitchens, hearing aids, thick accents, millisecond latency. Companions with long-term memory. Knows someone after six months the way a friend would.

Assistants inside messaging, connecting families across cities and countries. Health biomarkers from conversation. Cognitive decline, mood, breathing, loneliness, with research partners at Wharton, Toronto, and Oxford. PhD-level audio in production. Speech enhancement, VAD, diarization, TTS tuned for older voices. An agentic browser: booking appointments and getting things done online through conversation, no wrestling with websites. What we need; A cracked, AI-native engineer.

You use AI to multiply what you ship, and you've made yourself measurably faster with coding agents or your own setup. You've shipped something that runs on an LLM in production. Voice AI experience (real-time STT, TTS, speech-to-speech, telephony) is a big plus. You go above and beyond and can prove it: a side project with users, a repo people star, a hackathon you won, something you built over a weekend because you couldn't stop.

Show us. High agency. If something is broken and nobody owns it, you own it now. You'd rather be wrong fast than wait for a spec. Product taste. You care how it looks and feels, not just whether it works. Low ego, kind, hard-working. No politics. You care about the mission: older adults ageing well at home, work that matters to real families.

WHAT WE DON'T CARE ABOUT Where you went to university, or whether you did. Show us what you've built. Where you live. Overlap with London hours matters more than your postcode. Gender, age, background, or anything else that isn't the work. FOUNDING ENGINEER MEANS FOUNDING You will be integral to whether Harmonica succeeds. That's literal, and it comes with the freedom to match.

Set the engineering standards: how we test, review, deploy, and use agents as we grow. Shape the stack. A strong opinion backed by evidence changes how we build. Hire the team. You'll interview and help decide who joins after you. Own broadly today (voice, memory, messaging, biomarkers); take primary ownership of a major surface as we grow. Real equity and a real voice in product, engineering, and company decisions.

You'll learn more here in one year than in three at a corporate. That's a promise.