Senior Forge AI Engineer: Build Autonomous Agentic Platform

Norstella; Pierre (SD); Remote; USD 160,000 - 220,000; Full time

Job summary

Norstella is building Forge—a greenfield autonomous AI platform that replaces the SDLC with an agent chain, end-to-end. You’ll work across the full stack: orchestration, multi-agent coordination, MCP server design, RAG pipelines, and safety primitives to keep non-deterministic systems trustworthy in production.

The platform is live and shipping with hard problems in memory, trust, cost attribution, and regulatory compliance.

Qualifications

• 6+ years of software engineering experience, with production experience in agentic systems.
• Experience with at least one agentic framework and understanding of trade-offs.
• Hands-on experience authoring MCP servers or tool-use APIs.
• Experience with agent memory architectures and state persistence.
• Strong Python and/or TypeScript fundamentals.
• Proficient in prompt engineering, tool calling, and multi-step orchestration.
• Knowledge of RAG patterns, vector databases, and LLM evaluation in production.
• Familiarity with AI safety engineering, including OWASP LLM Top 10 and defenses.

Responsibilities

• Design and build agents across the Forge chain and set agentic design patterns.
• Own multi-agent coordination architecture, trust boundaries, and recovery.
• Build and evolve the MCP server layer with contracts and versioning.
• Design agent memory and state management across multi-step workflows.
• Own agent quality end-to-end with eval harnesses and benchmarks.
• Instrument observability: token tracing and cost attribution per run.
• Build safe-by-design primitives for least-privilege tool-use and PII handling.

Skills

Agentic systems

LLM frameworks

MCP servers

Tool-use APIs

Python/TypeScript

Prompt engineering

RAG patterns

AI safety

Tools

LangGraph

CrewAI

Autogen

Semantic Kernel

MCP

Job description

Norstella is building Forge—a greenfield autonomous AI platform that replaces the SDLC with an agent chain, end-to-end. You’ll work across the full stack: orchestration, multi-agent coordination, MCP server design, RAG pipelines, and safety primitives to keep non-deterministic systems trustworthy in production.

The platform is live and shipping with hard problems in memory, trust, cost attribution, and regulatory compliance.