Brief (a16z sr005)
Founding Engineer - AI Agent Infrastructure (San Francisco)
Brief (a16z sr005), San Francisco, California, United States, 94199
Brief is a product management agent swarm that gives AI coding tools product judgment.
We're a 2-person founding team using Cursor/Claude to ship at 10x velocity. Here's what we're building:
AI coding assistants made you 10x faster at writing code. But they can't answer product questions: Should we build this feature? (they don't know your strategy) Why did we reject real-time collaboration? (they have no memory) What do our users actually need? (they can't read research calls) Is this the right priority? (they don't understand your velocity or constraints)
We're building the missing product intelligence layer.
Instead of just making developers faster, we're enabling a new way of work that blends product, design, and engineering. When you ask Claude a question in your IDE, Brief's agent swarm runs in the background: StrategyAgent extracts your vision and goals from roadmaps and docs ResearchAgent synthesizes user needs from customer calls VelocityAgent calculates what your team can actually ship DecisionAgent remembers what you tried before and why it failed WorkAgent knows what's in flight and what's blocked
Your AI coding assistant now has product judgment, not just coding ability.
The Technical Challenge
This isn't just CRUD. You're building: Agent orchestration - State machines that coordinate multiple LLM agents with different specializations, handling failures and synthesizing results Real-time context synthesis - Event-driven architecture processing webhooks from GitHub, Linear, Notion, Slack, extracting semantic meaning from unstructured data Conversational interfaces - Chat UIs with branching logic, extraction agents, and background orchestration that feel natural but are technically complex MCP protocol implementation - We shipped production MCP integration 2 weeks after Anthropic launched the spec. You'll be extending it as the protocol evolves Multi-source intelligence - Combining signals from code (GitHub), work (Linear), strategy (Notion), and conversations (Slack) into coherent product context
What We've Shipped Conversational onboarding with state machines and extraction agents VelocityAgent that calculates team throughput from GitHub commits Decision Intelligence that remembers rejected features and why Real-time Work Pipeline synthesis across all your tools Strategic Context Engine that extracts vision and goals from documents MCP integration so Cursor/Claude can access all of this
About The Role
You'll own entire agent systems end-to-end: Build new agents - Design and implement specialized agents (FeatureAgent, TechDebtAgent, etc.) with LLM-powered reasoning Agent infrastructure - Improve orchestration, add monitoring, optimize context windows, handle edge cases Integration layer - OAuth flows, webhook handlers, MCP servers, API design Full-stack ownership - React/Next.js frontend, PostgreSQL schemas, background jobs, streaming responses Product collaboration - Talk to users weekly, understand their workflows, make product decisions about what agents to build next
About You Must Have: 4-8 years building web applications with end-to-end ownership Strong TypeScript fluency Experience with async/event-driven architecture AI-native mindset: Heavy Cursor/Claude/Copilot user who's felt the pain of AI tools lacking product context Experience integrating with third-party APIs (OAuth, webhooks)
Strongly Preferred: Built conversational UIs or chat interfaces with complex state Worked on developer tools or infrastructure products Experience with LLM APIs (Anthropic/OpenAI) and prompt engineering Next.js 15 and PostgreSQL expertise Understanding of product management workflows
Bonus: Contributed to open source developer tools Built MCP servers or Claude Desktop extensions Experience with Inngest or similar orchestration tools Shipped features that required product judgment, not just technical execution
Our Stack Frontend: Next.js 15, TypeScript, Tailwind CSS, Radix UI Backend: Next.js API routes, Inngest, PostgreSQL/Supabase AI/ML: Anthropic Claude, OpenAI GPT-4, Vercel AI SDK, MCP Integrations: GitHub, Linear, Slack, Notion, Fireflies, Fathom Infrastructure: Render, Supabase, Clerk, Sentry
Why You Should Apply You're building the future of product development - AI agents that have product judgment, not just coding ability Technical depth + product impact - Complex agent orchestration that directly affects how teams build products Extreme ownership - You'll design, build, and ship entire agent systems. Your code goes to production daily. Direct user feedback - Talk to users weekly. See your agents make real product decisions in their workflows. Shape the category - We're defining what product intelligence for AI coding means. You'll help figure it out.
What Won't Work: Need lots of structure and defined requirements Want to build it right before shipping Uncomfortable with AI coding tools Pure frontend or pure backend focus
Compensation Salary: $140k-180k (based on experience and location) Equity: 0.5-1.5% (early team member grant) Benefits: Unlimited PTO Visa Sponsorship: At this time, we cannot sponsor work visas. Applicants must be authorized to work in the US.
Work Environment In-person, in our San Francisco office. We're a high trust get-shit-done team, so locking in at home is fine when needed, but the industry and product move too fast for fully remote.
How to Apply Email hello@briefhq.ai with: Your GitHub profile Resume What product decision you wish your AI coding assistant could make (1-2 sentences)
We read every application and respond within a few days. We are not accepting unsolicited resumes or submissions from external recruiters or agencies for this role.
Brief is an equal opportunity employer committed to building a diverse team. We encourage applications from people of all backgrounds, including women, people of color, LGBTQ+ individuals, people with disabilities, and veterans. We believe diverse perspectives make us better at serving our customers.
We're a 2-person founding team using Cursor/Claude to ship at 10x velocity. Here's what we're building:
AI coding assistants made you 10x faster at writing code. But they can't answer product questions: Should we build this feature? (they don't know your strategy) Why did we reject real-time collaboration? (they have no memory) What do our users actually need? (they can't read research calls) Is this the right priority? (they don't understand your velocity or constraints)
We're building the missing product intelligence layer.
Instead of just making developers faster, we're enabling a new way of work that blends product, design, and engineering. When you ask Claude a question in your IDE, Brief's agent swarm runs in the background: StrategyAgent extracts your vision and goals from roadmaps and docs ResearchAgent synthesizes user needs from customer calls VelocityAgent calculates what your team can actually ship DecisionAgent remembers what you tried before and why it failed WorkAgent knows what's in flight and what's blocked
Your AI coding assistant now has product judgment, not just coding ability.
The Technical Challenge
This isn't just CRUD. You're building: Agent orchestration - State machines that coordinate multiple LLM agents with different specializations, handling failures and synthesizing results Real-time context synthesis - Event-driven architecture processing webhooks from GitHub, Linear, Notion, Slack, extracting semantic meaning from unstructured data Conversational interfaces - Chat UIs with branching logic, extraction agents, and background orchestration that feel natural but are technically complex MCP protocol implementation - We shipped production MCP integration 2 weeks after Anthropic launched the spec. You'll be extending it as the protocol evolves Multi-source intelligence - Combining signals from code (GitHub), work (Linear), strategy (Notion), and conversations (Slack) into coherent product context
What We've Shipped Conversational onboarding with state machines and extraction agents VelocityAgent that calculates team throughput from GitHub commits Decision Intelligence that remembers rejected features and why Real-time Work Pipeline synthesis across all your tools Strategic Context Engine that extracts vision and goals from documents MCP integration so Cursor/Claude can access all of this
About The Role
You'll own entire agent systems end-to-end: Build new agents - Design and implement specialized agents (FeatureAgent, TechDebtAgent, etc.) with LLM-powered reasoning Agent infrastructure - Improve orchestration, add monitoring, optimize context windows, handle edge cases Integration layer - OAuth flows, webhook handlers, MCP servers, API design Full-stack ownership - React/Next.js frontend, PostgreSQL schemas, background jobs, streaming responses Product collaboration - Talk to users weekly, understand their workflows, make product decisions about what agents to build next
About You Must Have: 4-8 years building web applications with end-to-end ownership Strong TypeScript fluency Experience with async/event-driven architecture AI-native mindset: Heavy Cursor/Claude/Copilot user who's felt the pain of AI tools lacking product context Experience integrating with third-party APIs (OAuth, webhooks)
Strongly Preferred: Built conversational UIs or chat interfaces with complex state Worked on developer tools or infrastructure products Experience with LLM APIs (Anthropic/OpenAI) and prompt engineering Next.js 15 and PostgreSQL expertise Understanding of product management workflows
Bonus: Contributed to open source developer tools Built MCP servers or Claude Desktop extensions Experience with Inngest or similar orchestration tools Shipped features that required product judgment, not just technical execution
Our Stack Frontend: Next.js 15, TypeScript, Tailwind CSS, Radix UI Backend: Next.js API routes, Inngest, PostgreSQL/Supabase AI/ML: Anthropic Claude, OpenAI GPT-4, Vercel AI SDK, MCP Integrations: GitHub, Linear, Slack, Notion, Fireflies, Fathom Infrastructure: Render, Supabase, Clerk, Sentry
Why You Should Apply You're building the future of product development - AI agents that have product judgment, not just coding ability Technical depth + product impact - Complex agent orchestration that directly affects how teams build products Extreme ownership - You'll design, build, and ship entire agent systems. Your code goes to production daily. Direct user feedback - Talk to users weekly. See your agents make real product decisions in their workflows. Shape the category - We're defining what product intelligence for AI coding means. You'll help figure it out.
What Won't Work: Need lots of structure and defined requirements Want to build it right before shipping Uncomfortable with AI coding tools Pure frontend or pure backend focus
Compensation Salary: $140k-180k (based on experience and location) Equity: 0.5-1.5% (early team member grant) Benefits: Unlimited PTO Visa Sponsorship: At this time, we cannot sponsor work visas. Applicants must be authorized to work in the US.
Work Environment In-person, in our San Francisco office. We're a high trust get-shit-done team, so locking in at home is fine when needed, but the industry and product move too fast for fully remote.
How to Apply Email hello@briefhq.ai with: Your GitHub profile Resume What product decision you wish your AI coding assistant could make (1-2 sentences)
We read every application and respond within a few days. We are not accepting unsolicited resumes or submissions from external recruiters or agencies for this role.
Brief is an equal opportunity employer committed to building a diverse team. We encourage applications from people of all backgrounds, including women, people of color, LGBTQ+ individuals, people with disabilities, and veterans. We believe diverse perspectives make us better at serving our customers.