Founding
AI Engineer
We are building AI that makes people better at relationships. The technical problem is harder than it sounds, and the engineer who solves it will have built something that genuinely matters.
Current AI fails in high-stakes human contexts not because it lacks capability. It fails because it was never trained for relational competence. That is what we are building.
THE PROBLEM
Why this is hard and why it matters
Models have seen knowledge, code, and engagement-optimized social content. They have not seen what healthy relating looks like over time. The result is AI that flatters, avoids friction, and optimizes for the feeling of being understood rather than the reality of growing.
Better Half is building the alternative: a system with a decision layer that determines when a user needs pushback, softening, or repair. Your job is to build the execution layer that turns those decisions into conversation that feels natural while actually serving the person having it.
WHAT YOU WILL BUILD
Six systems. One mission.
Affect analysis pipeline
Real-time emotion detection across text, voice, timing, and interaction patterns. Tracks escalation and de-escalation to feed the decision layer.
Prompt engineering
Translating high-level decisions into adaptive prompts that balance warmth with necessary friction, without breaking the experience.
Cloud infrastructure
AWS or GCP at scale, PostgreSQL and vector stores for memory, privacy-preserving pipelines with on-device computation where possible.
Memory systems
Storage and retrieval of relational context across sessions, consistency checking, and long-term modeling of user growth and relational capacity.
LLM orchestration
Multi-model coordination across Mistral, Claude, and others as needed. Latency optimization and graceful fallback strategies for real-time conversation.
Evaluation and monitoring
Eval harnesses that measure relational outcomes rather than fluency. Tracing, feedback loops, and A/B testing for relational interventions.
WHAT YOU WILL NEED
Required and valuable
REQUIRED
-
Production-grade Python with async and clean architecture
-
LLM integration: prompt engineering, constrained generation, orchestration
-
Transformer models and sentiment and emotion classifiers
-
Cloud infrastructure on AWS or GCP, PostgreSQL, vector stores
-
Eval and monitoring: harnesses, tracing, and feedback loops
-
Ability to implement specs from academic papers
-
Multimodal inputs
-
Comfort with 0-to-1 startup pace
ALSO VALUABLE
-
Utility theory and behavioral economics
-
Reinforcement learning
-
Game AI or dialogue systems experience
-
Rust
Show us
how you think.
Send your resume and something you have built to jobs@better-half.ai. A link to code, a project, a writeup. We read everything.