Staff Software Engineer, AI Agent Platform & Runtime

About us

Datasite is the essential infrastructure for M&A. Also comprising Grata, Blueflame AI, and Sherpany, we help dealbreakers find opportunities, manage transactions, and make key decisions.

Whatever your talents, you can find a place for them here. Get involved in some of the biggest business ventures today. Work with other gifted people around the world. Follow your ambitions, or discover a whole new path.

Can you help us simplify the complex, create great experiences, and deliver better outcomes? Then let's talk.

Job Description:Why Engineering at Datasite

Datasite's engineering organization builds the software the world's dealmakers depend on to find, evaluate, negotiate, and close transactions: secure virtual data rooms, AI-assisted diligence, deal sourcing, and private-company intelligence.

We're looking for a hands-on Staff Software Engineer to shape the foundational infrastructure for enterprise AI agents. You'll evolve our custom agent harness, build layered memory and secure isolated execution, and make sophisticated agent workflows dependable at production scale. This is an opportunity to solve challenging problems in distributed systems, applied AI, security, and platform engineering - not simply wrap APIs around models.

Help define how agents reason, plan, remember, use tools, delegate work, and recover from failures across complex, high-stakes business workflows. You'll own architecture decisions and collaborate across Engineering, Product, Machine Learning, and Security to deliver meaningful capabilities for dealmakers.

Location: We are hiring in New York City and Minneapolis. Datasite is a hybrid company, with in-office collaboration on Mondays, Tuesdays, and Thursdays. What You Will Do

• Lead the architecture and evolution of Datasite's custom agent harness, including reasoning and execution loops, planning, tool selection, state transitions, structured outputs, and runtime controls.

• Design layered agent memory spanning working context, episodic history, semantic knowledge, and durable workflow state; build retrieval, consolidation, compaction, provenance, retention, and access controls.

• Architect isolated, containerized agent execution environments for code, files, and tools, with ephemeral workloads, tenant boundaries, resource quotas, restricted networking, and secure credential handling.

• Build resilient agent orchestration: asynchronous tasks, specialized subagents, delegation, checkpoints, resumability, retries, failure recovery, and human approvals.

• Create secure, extensible tool integrations using MCP and related interfaces, with authentication, authorization, least-privilege permissions, and safeguards against prompt injection and unsafe actions.