SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Pencil is an advertising technology company building a generative AI-powered SaaS platform that redefines content creation for brands and creative professionals. The company's mission is to make AI the default in advertising without replacing human creativity, serving everyone from small businesses to large enterprises.
As Head of Agent Architecture, you will define and lead Pencil's technical vision for agent systems across the platform. This is a player-coach role combining hands-on architecture work with team leadership. You'll own the strategic direction for multi-agent workflows, task decomposition, state management, and tool-use patterns (RAG, API integration). You'll build and mentor a small team of Agent Architects, setting technical standards and processes that enable the platform to serve 1,000+ brands with unique voices simultaneously.
Key responsibilities include: defining the technical roadmap for agent architecture and securing leadership buy-in; hiring and managing Agent Architects with real-time feedback and development conversations; championing evaluation frameworks to measure agent performance systematically; owning reusable agents and prompt libraries that balance platform efficiency with bespoke client needs; partnering cross-functionally with AI Engineering, Product, and Sales to determine whether solutions are handled via prompting, RAG, or fine-tuning; and serving as the final technical escalation point for complex client deployments, personally leading discovery on high-stakes engagements and translating ambiguous briefs into buildable agent workflows.
You'll represent Pencil's agent technology directly to brand and agency stakeholders, decomposing client problems and explaining technical trade-offs (cost, latency, reliability) in plain language to non-technical executives. You'll also coach your team to develop this same client-facing problem-solving capability.
Required background: 8+ years of direct GenAI experience with 2+ years leading people or owning a function; deep technical fluency with LLMs (GPT-4, Claude, Gemini) and multimodal models; systems thinking about latent space, context windows, and agent composition; comfort with Python, JSON, and API documentation; demonstrated evaluation obsession (building benchmarking/A/B testing practices); proven leadership hiring and growing technical talent or owning roadmaps under stakeholder pressure; experience leading discovery with senior customers and turning vague briefs into scoped solutions; and ability to explain agent failure modes and architectural trade-offs in plain language to non-technical stakeholders.
Success is measured by: reducing failure rates in complex agent workflows; minimizing token usage and latency while improving output quality; achieving high automated-evaluation scores against brand guidelines; driving core agent adoption across the user base; delivering custom client workflows on time with strong adoption; building stakeholder trust through repeat engagement and clear communication; and growing a high-performing team.
