Senior AI / macOS Development Team Needed for Agentic Desktop AI (R&D / MVP)

The goal is to investigate whether large language models can become persistent desktop operators, capable of using existing software exactly as a human user would. *** We are looking for a highly capable software development team to assist with the research and development of an ambitious AI project. This is not a traditional web application or SaaS platform. It is an experimental R&D project exploring how modern large language models can become true desktop agents capable of operating a user's computer. Our internal team is already developing the overall product concept and system architecture. We are looking for an external development partner to help design and build an initial proof-of-concept and MVP that validates the technical assumptions behind the project. ## Project Overview The proposed application will run natively on macOS (phase 1). Its purpose is to allow a cloud-based large language model, such as OpenAI GPT, Llama, Claude, Gemini or similar, to interact with the user's computer in real time. Rather than relying exclusively on APIs or predefined integrations, the system should enable the AI to interact with existing software through the graphical user interface, using the same mechanisms available to a human user. Examples include: * Operating browsers * Using applications such as Slack, Photoshop, WhatsApp, Excel or Finder * Navigating websites * Completing multi-step workflows * Creating persistent AI agents capable of performing recurring tasks autonomously One example agent could: * Every morning collect data from multiple systems * Generate a report * Create accompanying graphics using Photoshop * Publish the report on social media without APIs except browser * Notify coworkers through Slack * Continue monitoring for follow-up tasks The long-term vision is a framework where users can create and manage multiple autonomous desktop agents that continuously perform work on their behalf. The purpose of this engagement is not to build that complete platform, but to determine whether the underlying technical approach is commercially and technically viable. ## Scope The engagement will focus on research and development together with implementation of an MVP. Possible work areas include: * Evaluating native macOS technologies for desktop control * Accessibility APIs * Screen understanding * Keyboard and mouse automation * Vision-based interaction * Agent architecture * Prompt engineering * Long-running autonomous workflows * Reliability and recovery from unexpected UI changes * Multi-agent orchestration * Overall software architecture The final scope will be refined together with the selected team. ## Desired Experience We are primarily interested in teams that have experience within one or more of the following: * macOS native development (Swift / Objective-C) * Accessibility APIs * Desktop automation * Computer vision * Human-computer interaction * AI agents * Large Language Models * OpenAI API * Claude API * Llama * Multi-agent systems * Browser automation * Robotics Process Automation (RPA) * Computer Use models * Electron or native desktop applications Experience with projects involving autonomous software agents is considered a significant advantage. ## What We Are Looking For This project is significantly more research-oriented than a typical software development assignment. We are looking for engineers who enjoy solving open-ended technical problems, evaluating competing architectural approaches, and building experimental systems to validate hypotheses. Strong communication skills are important. We expect technical discussions, architecture reviews, and collaborative problem solving throughout the engagement. ## Proposal If applying, please include: * A short introduction to your team * Relevant projects involving AI, automation or desktop software * Experience with macOS accessibility or operating system APIs * Your thoughts on the technical feasibility of this concept * How you would approach building a first MVP * Which AI models you would recommend using today and why * Any technical risks or assumptions you believe should be validated first Please also indicate your team's location, approximate size, and availability. We are looking for a long-term technical partner if the MVP proves successful.