Digital Solution Architect, flexible working

Solution Architect - (Google Cloud, Agentic GenAI, MCP) Role summary People Tech Group is seeking a hands-on Solution Architect to design and lead an industrial Integrity Operating Window (IOW) AI platform for a leading refinery. You will own the end-to-end architecture across GCP data services, event/middleware, agentic GenAI (MCP microservices per source), and the AI & knowledge base layer (RAG over standards like API 571/584 and SOPs) - delivering a secure, real-time, expert-in-the-loop system Location: India (preferred IST timezone) Function: Engineering/Architecture Type: Full-time Responsibilities: -  Own the reference architecture: Translate business & SME needs into a cloudnative, real-time architecture spanning on-prem connectors (PI/IOW/LIMS/APM/ERP/Logbook) → MCP adapter layer → event bus → agentic AI → dashboard & workflows  Design MCP layer (per-source microservices): Define APIs, schemas, and contracts for MCP-PI, MCP-IOW/LIMS, MCP-APM, MCP-ERP, MCP-Logbook; standardize ondemand aggregates (24h/7d/30d/90d), normalized events, retries, and backpressure  Evented core: Architect and tune Pub/Sub (or Kafka) topics for deviations/alarms/actions; ensure idempotency, ordered processing where required, and replay strategy  AI & knowledge base: Lead agentic GenAI design—Data Analysis Agent (tool use over MCPs), Knowledge Agent (RAG over standards like API 571/584, vendor manuals, SOPs), Expert-Feedback Agent (thumbs/annotations loop), Orchestrator (GPT-class LLM)  RAG & vector search: Choose and implement vector store (Cloud SQL + pgvector, Vertex AI Matching Engine, or Milvus); define chunking, embeddings, refresh cadence, grounding/citations  Security & governance by design: Enforce VPC/VPN/Interconnect, IAM (least privilege/service accounts), Cloud KMS (CMEK), Cloud Audit Logs, Data Catalog & Lineage, DLP & data quality rules, SSO/OIDC  Data services & middleware: Select and size Cloud Run/GKE for MCPs & BFF, Cloud Storage for docs/dumps, Memorystore for feature cache, optional BigQuery (analytical store) without coupling runtime  Integration workflows: Orchestrate Action Tracker ↔ ERP (SAP PM/Oracle EAM) via MCP-ERP; define SLAs, ownership, and auditability for IOW→RBI hand-o# ;s (GE APM)  Observability & SLOs: Establish metrics, traces, logs, and SLOs (latency, freshness, accuracy), plus runbooks for incident response  Cost & scale: Model infra + LLM cost envelopes, implement autoscaling, request/response token budgets, and caching strategies  Delivery leadership: Produce architecture docs, ADRs, sequence diagrams; guide developers, data/ML engineers, and MLOps; engage SMEs (operations, corrosion/RBI) for iterative validation Required qualifications  10+ years in distributed systems/data/AI architecture with at least 3+ years on GCP leading production solutions  Proven design of event-driven architectures with Google Pub/Sub (or Kafka), and microservices on Cloud Run or GKE  Strong with GCP data stack: Cloud Storage, Cloud SQL/Postgres (pgvector), Dataflow/Data Fusion (or equivalent ETL), Cloud Monitoring, Cloud Logging, Secret Manager, IAM, KMS, VPC  Hands-on GenAI + RAG: embeddings, vector search, grounding/citations, prompt tooling, and safety/guardrails; experience integrating OpenAI/Vertex AI (Gemini) or Anthropic models  MCP-style adapter design: built source-specific microservices/APIs that normalize schemas, perform windowed aggregates, and publish events  Integration exposure to OSIsoft PI/PI Web API, IOW/LIMS, GE APM (RBI), and SAP/Oracle ERP (REST, OData, queues, or middleware)  Practical security & compliance: IAM design, network segmentation, CMEK, auditability, lineage; understands safety-critical environments  Excellent stakeholder skills: can translate SME input (e.g. API 571/584, IOW practices) into technical decisions Nice-to-have (preferred)  LLMOps/MLOps: Vertex Pipelines, model registries, prompt/versioning, evals, AB tests, telemetry  Action management integrations (SAP PM/Maximo), Apigee/API gateway exposure  Time-series modeling (feature engineering, basic LSTM/TFT/Prophet, or XGBoost classifiers for risk scoring)  Graph/causal modeling for chain-link analysis (propagation across units)  Experience hardening BFF/API layers, RBAC in UI, and expert-feedback/annotation UX  Oil and Gas Industrial domain: Refining, IOW (knowledge on standards like API RP 571/584), corrosion loops, RBI workflows