Lead AI/ML Engineer, Sleep Chat

We're building an AI-powered sleep support chat that scales Hatch's sleep expertise to thousands of parents simultaneously. This isn't a wrapper around ChatGPT—it's a production system that needs to understand complex sleep scenarios, align with Hatch's sleep philosophy, handle multi-turn conversations about routines/alarms/cues, and integrate with our IoT device ecosystem.

You'll architect and build the NLP pipeline from first principles while coding alongside the team. This is hands-on technical work driving a launch that's critical to our business.

Week 1:

• Audit existing AWS Lambda prototype and identify technical gaps
• Review Hatch's sleep content library, device APIs, and customer conversation patterns
• Ship initial prompt improvements and establish baseline metrics for response quality

Within 3 Months:

• Own the build-out delivering Phase 2 of AI Sleep Chat for Q2 2026 launch
• Build evaluation system for measuring response accuracy against Hatch's sleep philosophy
• Partner with sleep consultants to tune AI responses and establish guardrails
• Instrument logging/monitoring for intent recognition, response quality, and conversation outcomes

Within 6 Months:

• Lead Natural Language Interface project: translate voice/text commands into device API calls
• Design and implement multi-agent architecture for command processing pipeline (intent → parameters → CMS lookup → device control)
• Build routine creation wizard using LLM-powered recommendations
• Establish AI/ML infrastructure patterns for Hatch (model evaluation, fine-tuning, deployment)

Technical Depth:

• 7+ years Python development, 2+ years building production NLP(natural language processing)/LLM systems
• Expert with LangChain, prompt engineering, and orchestrating multi-agent LLM workflows
• Experience with Claude/GPT-4 APIs, AWS Bedrock, or similar LLM platforms
• Proven track record shipping conversational AI products that handled real customer queries at scale

AI/ML Systems:

• Build evaluation frameworks for LLM output quality
• Experience fine-tuning models or implementing RAG systems for domain-specific knowledge
• Worked with intent classification, entity extraction, and multi-turn dialogue management
• Know when to use off-the-shelf models vs. custom training (and can explain the tradeoffs)

Production Engineering:

• Debug issues across LLM→API→device boundaries in production
• Experience with async Python, Lambda/serverless architectures, and real-time processing
• Built monitoring/observability for AI systems (hallucination detection, response quality degradation)
• Shipped features handling ambiguous requirements under aggressive timelines

Domain Translation:

• Can interview domain experts (sleep consultants) and translate knowledge into system constraints
• Experience working with non-technical stakeholders to define AI behavior and guardrails
• Comfortable explaining AI decisions to product/business teams without jargon

• Need extensive onboarding on LangChain or modern LLM tooling
• Prefer research over shipping production systems
• Haven't built conversational AI that real users relied on
• Looking for pure ML research role without engineering responsibilities

This isn't another chatbot. Parents use Hatch devices at 2am when their baby won't sleep. The AI needs to understand "my 8-month-old keeps waking up at 3am" and provide accurate, safe guidance aligned with sleep science. Bad advice impacts infant sleep and parental sanity. The technical challenges—intent mapping for complex device routines, guardrails ensuring safety, voice command robustness—are legitimately hard. If you want to build AI that actually helps people during their hardest parenting moments, this is it.