AI Engineer - Full Stack

What We're Building

Context Engineering

• Design and build the system that assembles structured context models from heterogeneous data sources (SEO tools, CRMs, enrichment platforms, social APIs, AI search engines) and passes them to LLMs for strategic output. The architecture of context assembly is the single biggest determinant of output quality. This is not prompt engineering. This is context engineering at a systems level. You decide what information goes in, how it's structured, how it's weighted, and how it evolves per customer over time.

LLM Orchestration

• Build the multi-step LLM pipelines that take raw data and produce structured strategic outputs: prioritized actions, content briefs, competitive analyses, lead scoring, signal detection. Each pipeline involves multiple LLM calls with different objectives, structured output parsing, validation layers, and fallback logic. You own the reliability and quality of every pipeline.

Learning Infrastructure

• Design the feedback mechanism that makes the system compound. When an action is executed and an outcome is recorded, the next intelligence cycle incorporates that result. Over weeks and months, the system calibrates to each customer's specific market reality. You architect how learning happens, how outcomes are attributed, and how historical patterns influence future recommendations. This is the difference between a static tool and an intelligence system.

AEO/GEO Intelligence

• Build the infrastructure for running systematic AEO experiments. Track what content structures get cited in AI-generated answers, across which engines, for which query types. Build the data pipeline that captures these patterns and turns them into actionable signals that feed back into the product. This is applied AI research with a direct product application.

Agent Design

• Design and build autonomous agents that handle specific workflows: research agents that analyze a company's digital presence in under 60 seconds, signal detection agents that identify strategic opportunities from raw data, content intelligence agents that generate briefs calibrated to a specific company's context. Each agent needs to be reliable, fast, and produce output that a founder would trust enough to act on.

Model Selection & Cost Optimization

• Determine which models to use for which tasks. Not everything needs the most expensive model. Not everything can run on the cheapest one. You'll build a routing layer that matches task complexity to model capability, manages cost per customer, and ensures output quality stays above the threshold where the product is useful.

Who This Is For

• You've built LLM-powered systems in production.
• Not wrappers around ChatGPT.
• Systems where the AI does real work and the output has to be reliable enough that someone makes decisions based on it.
• You've dealt with hallucination, structured output failures, context window limits, and the gap between "works in a demo" and "works at 2am when nobody is watching."
• You think about AI as an engineering discipline, not a prompting exercise.
• You care about evaluation frameworks, output validation, context architecture, and the systematic measurement of quality.
• You understand that the hard part is not getting an LLM to generate text.
• The hard part is getting it to generate the right text, consistently, at scale, for different customers with different contexts.
• You've built multi-step LLM pipelines.
• You know that chaining multiple LLM calls introduces compounding error rates and that the architecture of the chain matters more than any individual prompt.
• You've designed retry logic, validation gates, and fallback strategies for when a step in the chain fails.
• You've worked with multiple LLM providers and understand the tradeoffs.
• Latency vs quality vs cost.
• Structured output reliability across different models.
• Context window utilization strategies.
• You have opinions about when to use Claude vs GPT vs open-source, and those opinions are grounded in production experience, not blog posts.
• You can build the infrastructure around the AI.
• Data pipelines, API integrations, evaluation harnesses, monitoring dashboards, cost tracking.
• The AI is the core, but the engineering around the AI is what makes it production-grade.
• Specifically valuable experience:
• Production LLM applications (not prototypes, not hackathon projects)
• Multi-step agent architectures with reliability guarantees
• Context window engineering and structured output parsing
• LLM evaluation frameworks and quality measurement
• Cost optimization across multiple model providers
• Real-time or near-real-time AI pipelines
• Python and/or TypeScript in production AI systems

What You Get

• Founding equity.
• The AI layer is the core IP of the company.
• Your work is the product.
• Your equity reflects that.
• A research-grade problem in a production context.
• Context engineering for compound intelligence systems is at the frontier of applied AI.
• You're not fine-tuning a chatbot.
• You're building an intelligence system that gets measurably smarter per customer per week.
• The evaluation of "smarter" is itself an open problem that you'll define.
• Architectural ownership.
• No AI committee.
• No model review board.
• You decide the model strategy, the agent architecture, the evaluation framework, and the cost structure.
• You own the AI layer end to end.
• **Data that most AI engineers never get access to.**
• Real B2B market signals, real content performance data, real conversion outcomes, across multiple companies and verticals.
• The dataset you'll work with is the kind of thing that makes AI systems actually useful, not just impressive in demos.