Senior Context Engineer, MagicSchool AI

• Deep Systems & AI Experience: 7+ years building distributed systems with at least 3+ years in staff/senior roles. Hands-on experience building LLM applications, agentic systems, or context-heavy AI workflows with clear understanding of transformer architectures and attention mechanisms
• Context Engineering Expertise: Demonstrated experience managing context windows, building dynamic retrieval mechanisms, or designing context compaction strategies. Understanding of when context becomes a liability vs. an asset
• Technical Stack: Strong coding skills in TypeScript/Node.js with Python as a nice addition. Experience with our stack (TypeScript, Node.js, PostgreSQL, NextJS, Supabase) or similar. Familiarity with LLM APIs (OpenAI, Anthropic, etc.) and their context management patterns
• Information Architecture: Understanding of information retrieval, structured data representation, and strategies for organizing knowledge for AI consumption
• Educational Context Awareness: Understanding of or interest in how educational content is structured (standards, curricula, taxonomies), privacy requirements (FERPA/COPPA), and how context needs differ across teaching scenarios
• Leadership & Impact: Track record of architecting information systems, making high-leverage architectural decisions, and mentoring engineers on sophisticated technical concepts

Desirable

• As a Senior Context Engineer for AI Systems, you'll architect and optimize how MagicSchool's AI agents reason, remember, and operate within complex educational workflows
• You'll design context management systems that determine what information our agents see, how they maintain state across multi-turn interactions, and how they retrieve knowledge without overwhelming their attention budget—ensuring reliable, coherent AI assistance for educators
• In this role you'll implement the technical foundation of how AI agents manage their "mental workspace", and ensure our agentic capabilities remain accurate and focused in classroom scenarios
• Context Architecture & Token Optimization: Design and implement context curation systems for product features that determine what information enters each agent inference step, balancing comprehensiveness with the finite attention budget of LLMs to prevent context rot
• Task-Specific Context Management: Build memory compaction mechanisms and state management patterns that allow agents to maintain coherence within specific teaching workflows (lesson planning, differentiation, assessment creation)
• Context Monitoring & Improvement: Implement monitoring and evaluation for retrieval precision and reasoning coherence within your features. Collaborate with staff engineers and the evaluations team on measuring attention allocation and performance
• Just-in-Time Knowledge Retrieval: Build dynamic, runtime data fetching that enable agents to autonomously pull relevant curriculum content, student data, and educational resources exactly when needed, rather than pre-loading context with unnecessary information
• Tool & Integration Implementation: Build token-efficient tool APIs and retrieval layers for product teams to build context-aware features
• Product Collaboration: Partner with Product to translate educational workflows into optimal context configurations, understanding which information signals matter most for different teaching scenarios
• Technical Collaboration: Work with evaluations researchers, platform engineers, and others to implement memory modules, retrieval adapters, and human-in-the-loop correction systems within your team's applications
• Squad Technical Leadership: In collaboration with Product and peers, shape work for the team. Coordinate with other squads on requirements, patterns, and integrated projects
• Knowledge Sharing: Share context engineering insights with the team, contribute to engineering discussions on context management patterns, token budget awareness, and dynamic information curation