Core Architect & Integration Lead — Applied AI

zerohash's Applied AI team manages two fronts: an internal Intelligence Core (a system connectivity and control layer) and external agentic-commerce products. This role acts as the engineering backbone for both—building the access layer, audit logging, skills/agents, and system architecture. The Applied AI team builds the Intelligence Core itself. Engineering maintains platform accounts, event hooks, and retrieval layers; this role builds everything above them.

Connect company systems securely, and produce records of agent action that a third party can verify.

The internal build acts as a sandbox for agentic commerce products. Identity, authority, write permissions, and tamper-evident logging will be proven internally before powering multi-institutional agentic networks.

• Own the access layer (native connectors, MCP servers, Databricks US/EU) using short-lived tokens, least-privilege read-only access, and zero raw PII while enforcing data residency.
• Define a versioned schema for two signed record types of credentials and action records, in one namespace so each can be checked against the other.
• Establish build discipline early: deployment pipelines, authority policy as code, skill merge gates, and signed builds with SBOMs.
• Build the authority lease service with integrated kill switches.
• Capture events into append-only, hash-chained logs for tamper-evident and independently verifiable auditing.
• Migrate the team’s existing agents onto the Core, launch new per-department agents and lead security reviews for continuous read access.
• Extend the Core from read to bounded write: logged actions, idempotency keys, compensating actions, and multi-approver time-locks for anything that moves money.
• Design core primitives for cross-institution interoperability, wrapping actions in signed mandates, scoped authority, and attestation receipts compatible with standards like x402, AP2, and MPP.
• Design self-hosted, zero-egress private stacks when required by classification evidence.
• Evaluate and integrate state-of-the-art developments in agent frameworks, confidential inference, and zero-knowledge attestations.
• Maintain a security-first posture by threat-modeling prompt injection, credential exfiltration, and confused-deputy risks upfront.

• Senior, Python-first backend engineering with production experience in event-driven systems, API integration, and CI/CD.
• Agent-system experience — composing LLM agents (Claude, GPT, or comparable; Bedrock or comparable platforms) into standing systems — with the judgment to keep the models at the edges and the control plane deterministic.
• Demonstrate security-first habits in a regulated environment (MSB/VASP, bank, or comparable): least privilege and read-only as reflexes, not requirements to be reminded of.
• Working familiarity with capability-based authorization: attenuable tokens, workload identity, policy-as-code.

• Verifiable credentials, HTTP message signatures, or any of the emerging agent-payment mandate schemes (x402, AP2, MPP, Visa TAP, Mastercard Agent Pay).
• Self-hosted LLM stacks; confidential computing / trusted execution environments.
• A research habit and autonomy: this is a foundational seat on a new team, with wide scope and few hand-holds.

• Healthcare Insurance: zerohash covers roughly 100% of employee premiums as well as a portion of spouse/children (U.S. only)
• Vision & Dental Insurance (U.S. only)
• Chance to earn equity
• Maternity & Paternity leave (after 6 months)
• WeWork All Access Membership
• WFH Yearly Stipend
• L&D Yearly Stipend (after 6 months)

zerohash is the leading crypto and stablecoin infrastructure platform and the operating system for digital money, powering the next…