Ditto AI is building the agentic social network — and we are doing it as an AI-native company.
We’re not trying to replace human interaction. We’re trying to remove friction, reduce noise, and make introductions and connections more efficient, thoughtful, effective.
That requires solid engineering and pragmatic NLP work — not demos.
Role Overview
We’re hiring a Senior NLP Engineer (Agentic AI) to help design and build the language systems that power our agents.
You will work across model integration, conversation pipelines, agent orchestration, and language governance — with a focus on production reliability. Every profile runs as an AI agent that:
understands people beyond surface signals
uncovers hidden compatibility patterns
reduces noise and guessing
helps people connect smarter, not louder
This role is heavily hands-on. You’ll be building, shipping, and iterating — not just prototyping.
What You’ll Build
conversational systems that hold up under real usage
retrieval pipelines where accuracy actually matters
enrichment layers that structure messy data
durable context + memory systems
matchmaking and chatting logic that reveals hidden compatibility
runtime safety and observability
evaluation tied to real, measurable outcomes
learning loops that improve — without losing control
Everything must be observable, testable, and maintainable.
Must-Have Experience
We’re looking for an experienced builder.
7+ years of hands-on NLP engineering in production
2+ years building LLM-enabled systems in production
shipped conversational AI / multi-turn agent systems
LangGraph and/or LangChain used in production, not just prototypes
Proficient in TypeScript and Python
proficient using AI coding tools like Cursor and Claude Code
Retrieval systems + AI database using:
MongoDB
Redis
AI DBs — ideally Weaviate
Nice to Have
social / dating / recommendation platforms
trust & safety or moderation systems
Reinforcement learning — especially TD learning (super plus)
How We Work
We value engineers who:
think in systems, not just features
ship iteratively and measure results
design for failure modes, not just best cases
care about how systems behave in production
take ownership from design through deployment
