Applied AI Engineer

Position Summary

We’re looking for an Applied AI Engineer to help build the agentic infrastructure that powers automation across Dwelly. Our growth model is acquisition. Every agency we acquire brings new operational workflows, legacy processes, and manual work. If we manage each new agency manually, the economics of the model break. Our ability to scale depends on turning complex operational processes into reliable AI-driven and automated systems.

We already have successful AI-driven flows running in Tenant Find and Property Management. They work and deliver real business results. But they were built quickly and as separate systems.

The next challenge is to bring these approaches together: build reusable agentic infrastructure, establish a robust evaluation and observability layer, and create systems that allow us to automate new workflows quickly and reliably as Dwelly scales.

This is not an AI research role. You will be building AI systems that operate inside a real business and are expected to work reliably in production.

Key Responsibilities

1. Agentic infrastructure

• Design and build the core primitives behind our agentic systems, including memory, context management, tool-calling, orchestration, and feedback loops.
• Move us from one-off AI solutions toward reusable infrastructure where new workflows can be introduced quickly and with predictable reliability.

2. Evaluation & observability

• Build the evaluation framework that allows us to understand how our agents perform and why they succeed or fail.
• Make testing, tracing, debugging, and evaluating AI systems as fundamental to our engineering process as unit testing traditional software.
• Develop the systems that allow us to confidently assess an agentic workflow before deploying it into production.

3. Agentic development

• Push forward how we use LLMs to build software itself.
• Create workflows where coding agents and specialized subagents can explore repositories, implement changes, review architecture, check conventions, run evaluations, and iterate with minimal human coordination.
• Use agentic engineering workflows extensively in your own day-to-day development.

4. Orchestration

• Design systems where specialized agents, tools, deterministic software, and humans work together effectively.
• Understand when a problem should be solved with an LLM, when traditional software is the better solution, and when multiple coordinated agents can materially improve the outcome.

5. Operational automation

• Act as the bridge between complex operational workflows and engineering.
• Work closely with our operational and product teams to understand how work actually happens across acquired agencies, identify the highest-leverage opportunities for automation, and turn them into reliable production systems.

6. Architectural influence

• Help define the architectural patterns we use for agentic systems as the company scales.
• You won’t be joining as engineer #20 on an established AI platform. You will be an early core member of the team with significant influence over our technical approach, tooling, and engineering standards.

Qualifications and Preferred Background

• Strong software engineering background with experience independently delivering complex systems from idea through to production

• Hands-on experience building AI or agentic systems that have operated in production

• Experience dealing with real-world AI challenges such as reliability, latency, context management, failure modes, evaluation, and observability

• Strong understanding of agents, tool use, structured outputs, context management, orchestration, evaluation, and feedback loops

• Advanced practical use of modern LLMs — not simply using ChatGPT or Claude occasionally, but understanding how to build reliable workflows around probabilistic systems

• Experience using coding agents and agentic development workflows as part of your own engineering process

• Experience decomposing complex engineering tasks across agents or specialized subagents and designing mechanisms for them to verify and improve their outputs

• Strong architectural judgment and ability to decide when to use LLMs, deterministic software, human intervention, or a combination of approaches

• Experience with TypeScript / Node.js , or strong experience building AI systems in Python with the ability and willingness to apply that knowledge in a TypeScript-first environment

• High autonomy, ownership, and comfort making decisions under ambiguity

• Strong product judgment and ability to connect technical decisions to operational and business outcomes

• Strong communication skills and fluency in English

• Startup mentality: resilience, adaptability, and ability to thrive in a fast-paced environment

What We Don't Mean by “AI Experience”

For this role, AI experience does not mean:

• Using ChatGPT to generate snippets of code

• Adding a single LLM API call to an application

• Building a prompt wrapper and calling it an agent

• Blindly accepting whatever a coding agent generates

We are looking for engineers who think in terms of agents, tools, context, orchestration, evaluation, feedback loops, and human-machine collaboration — and who already apply these concepts in their day-to-day engineering work.

What Success Looks Like

Within your first 12 months:

• Reliability: We can deploy a new agentic workflow and understand how it will perform before it reaches production because we have robust evaluation and observability systems in place.

• Speed: Integrating processes from newly acquired agencies takes days rather than weeks because we have a standardized library of reusable agentic primitives.

• Efficiency: A significant amount of repetitive operational work has been automated, allowing our teams to focus on the areas where human judgment and interaction create the most value.

• Engineering leverage: Agentic development workflows materially increase the speed and quality at which our engineering team can build, test, and evolve software.

Technology Stack

TypeScript (Primary) · Node.js · PostgreSQL · LLM Orchestration · Agentic Workflows

Why This Might Not Be for You

This role is probably not the right fit if:

• You primarily want to research new models, read the latest papers, or train custom models. We are pragmatic builders rather than an AI research lab.

• You prefer working from a clearly defined backlog where someone else determines what should be built.

• You enjoy prototyping AI systems but are less interested in the less glamorous work required to make them reliable — evaluation, logging, error handling, observability, infrastructure, and operational edge cases.

• You see AI primarily as a feature rather than a new way of designing and building software systems.