🚀 SurfSense Call for Engineers

SurfSense is looking for exceptional engineers to help us build the future of enterprise AI search and collaboration.

Some facts about us:

• Open-source alternative to Glean with a rapidly growing community and real users
• Scaling from OSS to commercial product in 2025
• Building permission-aware RAG, real-time multiplayer AI, and deep agentic workflows
• Supporting 100+ LLMs and dozens of integrations
• Fully remote team, product-focused, shipping fast

If you want to own large parts of complex systems, move fast, and help define the architecture of an AI workspace used by real teams, this is for you.

👉 Applications are due by 8pm Pacific Time on Monday December 8, 2025. Interviews completed and decisions by December 12.

Projects described below.

Requirements

• You want high agency and ship fast without hand-holding
• You can take ideas to architecture to prototype to production
• Strong fullstack skills (React, Next.js, TypeScript, Python, FastAPI)
• Experience with AI-native tools (Cursor, Claude Code, model APIs)
• Knowledge of agents, RAG, embeddings, vector databases, and external integrations
• Strong debugging skills across distributed systems
• Comfortable working across frontend, backend, infrastructure, and AI workflows

Bonus skills:

• LangGraph, MCP servers
• Semantic search and vector database experience
• Realtime systems including websockets, PartyKit, CRDTs
• Y.js, ProseMirror, TipTap
• Docker and self-hosted deployments
• Enterprise SaaS permissions and ingestion pipelines
• OSS contributions

Projects for You

Below are the two major 2025 engineering tracks at SurfSense. You will own significant portions of one or both.

Project 1: Deep Agentic RAG and Realtime Agent UI

1. Introduction

We are building a highly optimized ReAct-style agent with long-running reasoning (Deep Agents) integrated directly into the SurfSense workspace.

This project includes:

• A deeply optimized agent engine for search and action execution
• Stateful long-running workflows using LangGraph
• A collaborative agent chat UI with realtime multiplayer features
• A modern developer-quality interface similar to high-end AI coding tools
• SurfSense MCP Servers that allow external tools to run agents and retrieve context from SurfSense
  
  

2. Requirements

Agent Engine

• Build agent loops using LangGraph
• Tool orchestration, memory, retrieval, and error handling
• Optimize semantic and hybrid search

UI and Realtime

• Realtime agent chat using websockets, PartyKit, and Vercel streaming protocol
• Comments, tagging, mentions, and notifications
• Frontend work with React, Next.js, TypeScript, Tailwind, shadcn, Framer

MCP Integration

• Build MCP Servers in FastAPI for agent execution and data retrieval
• Handle streaming, authentication, and permission-aware context

3. Non-Goals

• Agent marketplace
• Plugin ecosystem
• Working with non-permissioned data sources

4. Prior Art

• LangGraph Deep Agents (architecture inspiration)
• Anthropic MCP (server design patterns and tool interfaces)
• Cursor agent UI (interaction model and UX patterns)
• Liveblocks and PartyKit examples for multiplayer communication
  
  

Project 2: Multiplayer Document Editing 

1. Introduction

We are evolving SurfSense documents into a real-time collaborative workspace with deep AI integration.
Documents will become:

• Multiplayer and fully synchronized
• Versioned and auditable
• Commentable and taggable
• Searchable through SurfSense RAG
• Directly connected to our ingestion and connector pipelines
  
  

2. Requirements

Editor Architecture

• Implement multiplayer using Y.js and CRDT-based sync
• Build rich-text editing with TipTap and ProseMirror
• Add version control, diffs, and historical recoverability

Collaboration Layer

• Realtime comments, tagging, notifications
• Permission-aware access control at document and block level
  
  

Search and RAG Integration

• Make document changes searchable instantly
• Optimize RAG pipelines for incremental updates and permission filtering
  
  

Connector and Ingestion Work

• Improve ingestion pipelines using LlamaCloud, Unstructured, and Docling
• Strengthen connectors including Google Workspace, Microsoft 365, and databases

3. Non-Goals

• Complete office suite
• Binary file editing
• Offline editing
  
  

4. Prior Art

• Google Docs for collaborative editing behavior
• Notion for comments, tagging, and document interaction patterns
• Linear for clean, structured collaboration UI
• Yjs collaborative demos and CRDT synchronization patterns
• VS Code Live Share for multiplayer reliability concepts

surfsense.com | github.com/MODSetter/SurfSense