Manager - Customer Interaction Suite

Tata Communications; Scotland; On-site; GBP 110,000 - 190,000; Full time

Job summary

Tata Communications Limited is seeking an AI OS Architect to turn complex customer operations into a clean ontological model, wire data and systems into it, and orchestrate LLM-driven agents and workflows with governance and cost discipline.

The role requires experience in deploying an enterprise AI platform, strong data engineering capabilities, and the ability to lead on-site customer engagements and design reusable architecture for multiple deployments.

Qualifications

• Must-have enterprise AI OS architecture experience with deployed platform
• Ontology-first thinking modelling enterprise objects, actions and processes
• Deep data engineering pipelines, lakehouse and data lineage
• LLM/agent orchestration, context engineering and tool use
• Enterprise governance, security IAM, and GDPR/SOC2 awareness
• Forward-deployed temperament: travel, ambiguity, client-facing

Responsibilities

• Define enterprise ontology and data architecture for reusable modelling
• Lead customer discovery and map objects, relationships, and processes
• Architect data foundations from ops systems to governed data layer with lineage
• Design reusable object/action/process models across AI workflows
• Architect AI platform, RAG, agent orchestration, and governance
• Lead 1-2 major customer engagements on-site through delivery and launch
• Develop quick, value-proving thin slices and capture field learnings for productFeedback

Skills

Enterprise AI OS architecture

Ontology-first modelling

Data pipelines & lakehouse

LLM orchestration & agents

Security & governance

Forward-deployed delivery

Executive-level communication

Tools

SQL

PySpark

Job description

About The Company

Tata Communications Redefines Connectivity with Innovation and IntelligenceDriving the next level of intelligence powered by Cloud, Mobility, Internet of Things, Collaboration, Security, Media services and Network services, we at Tata Communications are envisaging a New World of Communications

About The Role

Tata Comm AI OS proved the model with Context Graph and AIP an ontology-first architecture where every data asset, action, and process is modelled once and reused everywhere.

You are not an internal software architect. You are the AI OS architect on the customer s floor who turns their messiest operational reality into a clean ontology, wires their data and systems into it, and orchestrates LLM-driven agents and workflows on top, all within governance, security, and cost discipline. If you have lived this, or built the same discipline with other enterprise AI platforms (Palantir, Databricks, Snowflake, AWS/Azure/GCP AI stacks, or an in-house equivalent), I want to talk.

Key Responsibilities

Enterprise Ontology & Data Architecture (the core of the role)

Lead discovery at customer sites: map the enterprise s objects (customers, orders, assets, claims, equipment), their relationships, actions, and processes then model them into a shared, reusable ontology, not a pile of point solutions.

Architect the data foundation: pipelines from operational systems, ERP/CRM, telemetry, documents and knowledge bases into a governed data layer (zero-copy / lakehouse patterns), with lineage, quality rules, and access controls that hold up to audit.

Design for reuse: every object, action, and process modelled once permissioned, versioned, and composable across all downstream AI workflows.

AI Platform & Agent Architecture

Design the AI layer on the enterprise OS: RAG over enterprise knowledge, context engineering, LLM/agent orchestration, tool-calling, human-in-the-loop fallbacks, and evaluation loops using no-code, low-code and pro-code paths as the situation demands.

Architect agentic workflows (autonomous, supervised, and human-in-the-loop)

Own model governance: model selection and cost per action, prompt/version control, guardrails, safety and bias checks, drift monitoring, and audit trails.

Forward-Deployed Delivery

Be the technical lead embedded in 1-2 major customer engagements at a time: run joint design workshops, build the reference architecture, and stay on-site through launch wrangling real, messy, production data alongside customer teams.

Build the thin slice that proves the model early (one decision, one agent, one happy path), then expand demonstrating value in weeks, not quarters.

Feed field learnings back: what patterns win in the market becomes input to our product and portfolio teams.

Must-Have (Non-Negotiable)

• Relevant AI OS architecture experience mandatory. You must have architected and shipped an enterprise AI platform deployment in a forward-deployed or equivalent embedded setting: at Palantir, Databricks, Snowflake, AWS/Azure/GCP AI platforms, or a comparable in-house enterprise AI platform. Generic ML/DS experience without enterprise-deployment depth will not clear the bar.


• Ontology-first thinking demonstrated ability to model an enterprise s objects, actions and processes as a reusable semantic layer, with metadata, permissions and data structures kept consistent, secure, and scalable Ontology.


• Deep data engineering pipelines, lakehouse/data-platform architecture, streaming and batch, data quality and lineage, zero-copy patterns; SQL, PySpark or equivalent fluency.


• LLM/agentic AI depth RAG architectures, context engineering, prompt and agent orchestration, tool use, evals, guardrails, and cost/latency tuning; hands-on with at least one major LLM stack.


• Enterprise governance & security IAM, row/object-level permissions, model governance, compliance (SOC2/ISO27001/GDPR ), and speaking that language to CTOs and CISOs.


• Forward-deployed temperament you are measured by customer outcomes, you travel, you tolerate ambiguity, and you can explain architecture to a COO and run a terminal in the same afternoon. Excellent written documentation and client-facing communication.

What Success Looks Like (KPIs)

Customer engagements that reach production: number of deployments delivered from architecture to go-live at the customer site.

Time-to-first-value: weeks from kickoff to a production thin slice per engagement.

Ontology/platform reuse: patterns and assets built once, reused across multiple customers.

Deal technical win rate and accuracy of BoM/effort estimates versus contracted scope.

Field feedback that demonstrably shaped product roadmap.