The Agentic AI Architect designs, governs, and enables agentic AI systemsAI solutions that can plan, reason, call tools/APIs, interact with enterprise systems, and execute workflows under strict security, safety, and compliance guardrails.
This role is accountable for the reference architecture, design standards, and productionization of agentic AI capabilities across the organization, ensuring solutions are :
– Secure-by-design and audit-ready
– Reliable and operationally supportable (247) with clear run-state ownership
– Cost-controlled with measurable outcomes and usage governance
– Safe and aligned through policy constraints, human oversight, and robust evaluation
The role partners closely with security/GRC, data engineering, platform engineering, IT operations, and product teams to deliver enterprise-grade AI agents that improve productivity, automate decisions, and orchestrate actions across systems.
Role Scope & Typical Use Cases :
Agentic AI solutions may include :
– IT Ops / Service Desk agents : guided triage, knowledge retrieval, assisted remediation, change validation
– Business operations agents : process orchestration, approvals, exception handling, document automation
– Engineering agents : code review assistance, runbook execution with approvals, infrastructure change suggestions
– Knowledge agents : enterprise search, policy Q&A, summarization with citation/traceability
– Multi-agent workflows : plannerexecutor patterns, supervisory agents, task decomposition across specialist agents
Key Responsibilities :
1) Agentic AI Architecture Strategy & Governance :
– Define enterprise agentic AI target architecture, including :
a. Agent patterns (single-agent, plannerexecutor, supervisor, multi-agent)
b. Orchestration and tool-use architecture
c. Policy enforcement points and approval flows
– Establish reference architectures, templates, and reusable components for agent solutions.
– Lead/participate in Architecture Review Boards (ARB) for agentic AI initiatives :
a. Approve design artifacts (HLD/LLD) and control mappings
b. Manage exceptions with documented risk acceptance
– Define the agent readiness checklist for production go-live (safety, security, monitoring, fallbacks).
2) Agent Orchestration & Workflow Engineering (Hands-On) :
– Architect and guide implementation of agent orchestration :
a. Task planning, decomposition, tool routing
b. Memory and context management (short/long-term; scoped and governed)
c. Stateful vs stateless execution patterns
– Define integration patterns to enterprise workflows :
a. BPM/workflow engines and orchestrators
b. Event-driven triggers and queue-based execution
c. Human-in-the-loop approvals for high-risk actions
– Ensure orchestration meets enterprise requirements :
a. Idempotency, retries, rate limits, timeouts
b. Concurrency control and safe execution boundaries (blast radius limits)
3) Tooling & Integration Architecture (APIs, Systems, Knowledge)
– Define secure tool-use patterns for agents :
a. API gateway integration, service tokens, scoped permissions
b. Read vs write action separation and approval gates
– Enable enterprise knowledge grounding :
a. Retrieval-Augmented Generation (RAG) architecture
b. Knowledge indexing, metadata, access control, and freshness rules
– Ensure agents integrate safely with :
a. ITSM (e.g., ServiceNow), monitoring systems, CMDB
b. Identity, access, and privileged systems (PAM)
c. Enterprise apps (ERP/CRM), document repositories, messaging tools
– Define contract-first integration (schemas, validation, error handling, audit logs).
4) Safety, Guardrails, and Responsible Agent Design
– Architect guardrails to prevent harmful/incorrect actions :
a. Policy constraints (what an agent can/cannot do)
b. Context filtering and sensitive data redaction
c. Output validation and schema enforcement
d. Tool-use allowlists/denylists and safe action sets
– Design human oversight patterns :
o Approval workflows for privileged actions
o Step-up verification for high-impact operations
– Establish misuse protections :
a. Prompt injection defenses (input sanitization, system boundaries)
b. Data leakage prevention (DLP alignment)
c. Model behavior controls and fallback behaviors
5) Security, Compliance & Audit Readiness (Regulated Environments)
– Embed security-by-design into agent ecosystems :
a. IAM/RBAC, least privilege, separation of duties
b. Secrets management, encryption, key management
c. Network segmentation and secure connectivity
– Ensure audit evidence readiness :
a. Immutable logs of prompts, tool calls, actions, approvals, and outcomes
b. Traceability from policy control implementation evidence
c. Retention, legal hold, and data governance alignment
– Partner with Security/GRC to define and maintain :
a. Control mappings, standard operating procedures, and risk assessments
6) LLMOps / AgentOps (Productionization & Run-State) :
– Define and implement AgentOps practices :
a. CI/CD for prompts, tools, policies, and orchestration logic
b. Versioning, approvals, rollback, and release gates
c. Environment separation (dev/test/prod) and change governance
– Establish operational monitoring and observability :
a. Latency, tool-call success rate, failure modes
b. Hallucination/incorrect-action detection signals
c. Cost and token consumption monitoring
d. SLA/SLO definition for critical agents
– Drive reliability engineering :
a. Graceful degradation when tools fail
b. Safe fallbacks (escalate to human, read-only mode)
c. Rate limiting and circuit breakers
7) Evaluation, Quality Engineering & Risk Controls :
– Define agent evaluation frameworks :
a. Offline evaluation sets (test cases, golden answers)
b. Scenario-based testing for edge cases and adversarial prompts
c. Regression tests for prompts and tool policies
– Implement quality gates for go-live :
a. Accuracy, safety, security, and reliability thresholds
b. Approval matrices and sign-offs
– Establish ongoing measurement :
a. Task success rate, human override rate, incident reduction
b. User satisfaction, time saved, compliance findings
8) Platform & Vendor Strategy
– Define standards for :
a. Model selection (commercial/open models), hosting, data boundaries
b. Vector stores, knowledge stores, and indexing approaches
c. Orchestration frameworks and automation integration
– Evaluate vendors and tools; support procurement with :
a. RFP inputs, TCO, risk posture, and architecture fitment
– Guide SIs/partners and internal teams with architecture assurance and technical leadership.
Are you interested in this position?
Apply by clicking on the “Apply Now” button below!
#AlbionarcJobs#FintechJobs
#AsiaJobs#MiddleEastCareers
#TechTalent#FintechRecruitment
#FinanceOpportunities#
