Applied AI Engineer

About the Role

We’re looking for a self-starter who loves building new products in an iterative, fast-moving environment. As a Founding AI Engineer , you’ll report directly to the cofounders and work closely with product and engineering. You’ll bring our smartest matchmaking AI to life, design chat agents that feel human, and create internal tools that other AIs use to reason, retrieve, and act. This is an early, high-ownership role (<10 people on the team) where your decisions will define our agentic system’s foundations.

In this role, you will:

• Ship agentic matchmaking from research to production—own the end-to-end loop (retrieval, reasoning, tool use, safety) and drive measurable accuracy improvements.

• Build a prompt & model evaluation harness (offline + online) to compare prompts/models/policies, support A/B testing, and enable fast iteration.

• Optimize AI chat systems for lower latency , higher perceived “human-likeness,” and more consistent outcomes across providers.

• Design and maintain context engineering pipelines (RAG, memory, summarization, compression, grounding) for conversations and matchmaking.

• Stand up observability for agents (traces, costs, failures, hallucinations, guardrails) and create dashboards that guide product decisions.

• Collaborate daily with the cofounders and product to translate user problems into agent behaviors, experiments, and shipped features.

• Write clear, maintainable code; create small internal tools and SDKs other engineers (and AIs) will use.

Your background looks something like:

• 2–4+ years of relevant experience or a standout personal portfolio of agents/LLM apps—show us what you’ve built (GitHub, demos, write-ups).

• Strong programming foundations (data structures, algorithms, testing, profiling).

• TypeScript (product code, tools, services) and Python (model ops, evals, data) proficiency.

• Experience building with multiple LLM providers and tool-calling/function-calling; comfortable swapping models and orchestrating fallbacks.

• Hands-on with RAG (indexing, chunking, embeddings, reranking) and context engineering for reliability and cost/latency trade-offs.

• Practical prompt engineering and prompt libraries; can reason about failure modes and systematically improve prompts/policies.

• Ability to define metrics/KPIs (accuracy, latency, cost, safety), run A/B tests, and loop in human feedback for quality.

• Comfortable with MongoDB in production; familiarity with vector databases (e.g., pgvector/Redis/Pinecone/Weaviate) is a plus.

• Extra plusses (the more the better): MCP (Model Context Protocol) , agent frameworks (LangGraph/CrewAI/Assistants), LLM observability/evals (e.g., Langfuse/Promptfoo/Ragas/TruLens), retrieval & embeddings know-how, safety/guardrails/red-teaming.

• Builder’s mindset: thrives with ambiguity, ships quickly, debugs systematically, and sweats the user experience.

Location: SF (onsite only)

About Ditto

Ditto is reimagining how people meet — starting with dating. We’re building the first fully agentic social platform where AI does the heavy lifting: understanding your preferences, finding compatible matches, and even setting up real-life dates.

Our cofounders dropped out of UC Berkeley in their freshman year to build this vision. Since then, Ditto has gone viral across campuses, set up tens of thousands of real dates, and attracted funding from Google and top-tier VCs, along with brilliant engineers and researchers from MIT, Stanford, Berkeley, and DeepMind.

Dating is just the beginning, we are gonna disrupt the entire social scene. If that sounds interesting, come talk to us.