Project: Merva (Intelligent Tutoring System) Location: Baltimore, MD (Hybrid/Remote) Duration: 12 Weeks (Summer 2026)

About Merva

Merva is an AI-native Intelligent Tutoring System (ITS) designed to achieve 100% automation in mathematics instruction. Merva operates on a rigorous "Teacher-Architect" philosophy, providing tailored interventions for teacher and student needs. The system will deploy a multi-agent AI architecture—hosted on local high-performance hardware—to generate scaffolded lessons, dynamically parameterize problems, and provide affective feedback based on cognitive load theory.

The Role

We are seeking a Junior Software Developer & AI Researcher for a 12-week summer sprint. You will work directly with the Lead Architect to build the core orchestration middleware of the Merva application.

An ideal candidate is an upper year or recent graduate with a degree in math + computer science.

While the Lead Architect will define the pedagogical logic, knowledge graph structures, and math curriculum, your primary focus will be engineering the multi-agent routing system, ensuring seamless communication between distinct AI agents and building the API bridges to the mathematical database.

Key Responsibilities

• Multi-Agent Orchestration: Design and implement a robust state-machine utilizing frameworks like LangGraph or AutoGen to govern interactions between specialized AI agents.
• Local Inference Integration: Assist in configuring and optimizing local LLM deployment (e.g., via vLLM, llama.cpp) for high-performance eGPU/workstation environments.
• Knowledge Graph API: Build the connective tissue between the multi-agent swarm and a deterministic graph database, allowing the AI to accurately parse prerequisite sub-graphs.
• Tool Use & Validation Failsafes: Integrate symbolic math libraries (like SymPy) and secure Python sandboxes so the AI can test, execute, and mathematically validate the problems and LaTeX strings it generates before presenting them to the student.
• Error Handling: Develop strict fallback protocols to ensure the system gracefully recovers from AI hallucinations or logic loops without breaking the student-facing instructional flow.

Required Qualifications

• Technical Background: Currently pursuing or recently completed a degree in Computer Science, Applied Mathematics, Computational Engineering, or a closely related field.
• LLM Orchestration: Hands-on experience building custom applications with large language models, specifically using agentic frameworks (LangChain, LangGraph, AutoGen).
• Mathematical Foundation: Strong background knowledge of undergraduate mathematics and AI.
• Systems Programming: High proficiency in Python and experience with asynchronous programming and API development.

Bonus Skills

• Experience with knowledge space theory or graph databases.
• Familiarity with open-source local inference engines and hardware optimization (VRAM management, quantization).
• Understanding of learning science principles, such as cognitive load theory.

Pay: $25.00 - $35.00 per hour

Work Location: Hybrid remote in Baltimore, MD 21201