Founding LLM & Backend Engineer Realtime Copilot (Prompt Systems, Tooling, Infra)
Type: Full-time, player-coach (hands-on IC now; grow a small team over time)
Why this role exists
Our AI companion needs to listen, reason, and respond in-the-momentwith empathy and precisionwhile staying private and reliable. You'll architect the LLM system that powers meeting-time guidance (micro-coaching, summaries, strategy suggestions), build the backend that makes it fast and safe, and define prompt/eval tooling that scales. You'll be equal parts prompt engineer, systems engineer, and product thinker.
What you'll do
- Own the conversational engine: tool-calling/function-calling, state machines, guardrails, JSON contracts, and prompt libraries for low-latency, high-reliability responses in live calls.
- Architect the realtime backend: streaming APIs (WebSocket/WebRTC), event buses, feature flags, rate limiting, and observability (traces, evals, red-team metrics).
- Build retrieval & memory: user-controlled personal context with strict privacy (PII redaction, scoped embeddings, data residency options).
- Evaluation & safety: automatic evals (factuality, toxicity, coaching quality), hallucination controls, regression suites, and human-in-the-loop review pipelines.
- Latency & cost engineering: prompt packing, caching, structured outputs, speculative decoding/streaming, and smart fallback policies.
- Voice & meeting UX plumbing: integrate ASR/TTS, diarization, and whisper-coach UI surfaces inside the extension without hijacking the conversation.
- Player-coach: define a 12-month platform roadmap, vendor selection (model/ASR/TTS), and recruit/mentor 13 engineers.
You'll thrive here if you have
- 510+ years building production backends and conversational systems; you've shipped LLM features to real users.
- Strong with TypeScript/Node (or Python) services, streaming patterns, and cloud infra (GCP/AWS).
- Experience with prompt engineering at scale: tool schemas, few-shot design, system prompts, eval harnesses, and telemetry-driven iteration.
- Familiar with multimodal inputs (audio/text; EEG context a plus), RAG, privacy & governance, and safe-completion strategies.
- Bonus: Chrome extension bridges, WebAssembly, vector DBs, on-device/edge inference, realtime voice agents.
About Myndland
We envision a workplace where calm is the foundation of strong relationships, trust, and influence. Our brainwave AI workplace companion helps professionals regulate emotions and communicate effectivelyturning stressful moments into opportunities for connection. We bridge neuroscience and AI to make people instantly likable, understood, and influentialby staying authentically calm and adaptive.
How we work
- Product > papers: we care about user outcomes and scientific integrityboth.
- Hands-on, high-agency: everyone ships. You'll own roadmap slices end-to-end.
- Privacy by design: encryption, minimization, and clear user control are table stakes.
- Iterate with users: tight loops, live telemetry, fast A/Bs.
What We Offer
- Collaborative, innovation-focused environment that values technical excellence and creative problem-solving
- Competitive compensation package in a high-potential startup
- Comprehensive benefits including health insurance and wellness programs
- Professional development opportunities, conference attendance, and access to cutting-edge research
- Flexible work arrangements including remote work options in a fast-moving startup culture at the intersection of AI and neuroscience
We are committed to fostering a diverse and inclusive workplace. We encourage applications from candidates of all backgrounds.
