AI Developer - LLM Features & AI Systems

STAN AI•Toronto, ON

1d•$85,000 - $120,000•Onsite

About The Position

STAN AI is the AI assistant for HOA and condo association property managers. We work with 20,000+ communities representing 2.5 million homes across North America, and we just closed our Series A. Our platform integrates natively with every major property management software in the category — Vantaca, CINC, Enumerate, eUnify, VMS, Caliber/FrontSteps, Buildium — which is why management companies are switching to us faster than this market has ever moved. We're looking for an AI Developer to build the AI backbone of our product — retrieval-augmented generation pipelines, multi-step agent workflows, embedding systems, and LLM integrations that property managers rely on daily. You'll work directly with product and engineering to ship AI features end-to-end: designing vector search strategies, building agent loops, evaluating model quality, and shipping systems that actually work in production. You won't just execute tickets — you'll bring a point of view on embedding models, chunking strategies, reranking approaches, and the real tradeoffs between quality, latency, and cost.

Requirements

Hands-on experience building RAG systems in production (chunking, embedding, retrieval, reranking)
Real experience with embedding models (OpenAI, Cohere, or open-source) and vector databases (Pinecone, Weaviate, Chroma, or similar)
Experience building agent loops or multi-step reasoning systems (tool use, memory patterns, error handling)
Familiarity with Claude API and/or OpenAI API — prompt design, function calling, streaming
Strong TypeScript and Python — you write clean, maintainable, well-tested code
Understanding of LLM limitations: hallucination, context windows, latency, inference cost, and real-world tradeoffs

Nice To Haves

Experience with Claude Code or AI-assisted development workflows
Knowledge of LLM evaluation frameworks (RAGAS, custom metrics, semantic similarity scoring)
Side projects or portfolio demonstrating real AI work (not tutorials) — GitHub, demos, case studies
Hands-on experience with orchestration frameworks (LangChain, LlamaIndex, or equivalent)
Experience with multi-modal inputs or structured output extraction (JSON mode, schema validation)
Background shipping AI features in a production SaaS environment (not just experiments)
Familiarity with Stan AI stack: Node.js, TypeScript, MongoDB, AWS
Understanding of prompt engineering, few-shot learning, and in-context optimization
Fine-tuning or RLHF experience
Contributions to open-source AI projects
Experience in PropTech, FinTech, or operations software
Knowledge of prompt injection risks and AI safety patterns
Familiarity with vector database administration (indexing, cost optimization, scaling)

Responsibilities

Design and build retrieval-augmented generation systems. Own chunking strategy, embedding selection, retrieval optimization, and reranking.
Implement and manage vector search infrastructure (Pinecone, Weaviate, or similar). Integrate embeddings with our MongoDB core data layer.
Build multi-step agent loops with tool use, memory, planning, and guardrails. Handle edge cases like hallucination, context limits, and reasoning failures.
Integrate Claude and OpenAI APIs using orchestration frameworks (LangChain, LlamaIndex, or equivalent). Manage prompts, context windows, streaming, function calling, and tool use.
Build evaluation pipelines to measure LLM output quality. Iterate on prompts, retrieval strategies, and model choices based on real data.
Use Claude Code and modern AI-assisted development as part of your workflow. Help the team ship faster with AI tools.
Work with product to scope AI features and advise on feasibility. Help set patterns and best practices as the AI feature set grows.