About the role

The TDA is 7SG's lead post-sales architect on a partner engagement. They design, build, and are at times responsible for the operations of production architecture. On most engagements they are hands-on: writing the HLD, most of the LLD, making the infrastructure and data platform calls, defining AI Execution layer configurations, and specifying integration patterns. On engagements a GSI partner leads, they review and approve the design before anyone builds. Either way, the technical outcome is theirs.

The primary architecture is the service provider one. A telco or neocloud operates the platform as a multi-tenant environment and sells capacity on to its own enterprise and government customers. That design has to survive a proving ground build-out, IaaS/PaaS portal deployment, infrastructure provisioning, and contracted operations until operational handover to the partner's own operations team. The secondary case is the enterprise running AI across a heterogeneous estate it already has. The TDA architects for both and leads with the first.

Three domains have to work together:

1. Infrastructure: private AI stacks (GPU clusters, switching fabrics, Kubernetes platforms, storage), hyperscaler IaaS/PaaS, neocloud and sovereign environments, SASE, BC/DR.
2. Data and intelligence: intelligence products (structured, vector, graph), data connectivity, pipelines, memory and telemetry, feedback loops.
3. AI Execution and application: identity and execution envelope, security and compliance controls, model and tools management, workload management, agent and workflow design.

The TDA works these domains across on-prem, hyperscaler, and sovereign environments, and knows what is different about each.

Success means no production incident traces back to a design risk nobody identified. Every architecture trades something: resilience against cost, latency against complexity, control against velocity. The job is to make those tradeoffs explicit, document them with agreed tolerances, and make sure monitoring fires before conditions reach the bounds.

What you'll do

Architecture and design

• Produce and own the technical architecture for partner engagements: HLDs, LLDs, implementation plans, acceptance criteria, test plans

• Align to AI Execution layer configurations: identity mediation, model routing and endpoint governance, workload profiling and placement, security and compliance postures, telemetry requirements

• Align to intelligence product architectures: data connectivity from core systems, vector, graph, and structured product schemas, pipeline patterns (CDC, streaming, micro-batch), cross-environment replication

• Design infrastructure for private AI deployments: Kubernetes platform layout, accelerator server sizing, switching fabric design (Spectrum-X, UEC), storage subsystems, integration with infrastructure provisioning software

• Design for multi-tenancy: tenant isolation, capacity provisioning through the customer portal, quota and chargeback models, and the operational handover the partner's team inherits

• Specify multi-environment routing: which workloads run on SaaS endpoints, hyperscaler endpoints, or on-prem and sovereign runtimes, based on data classification, cost, latency, and compliance constraints

• Define operational requirements: support severity models, workload profiling baselines, SLO enforcement, CVE remediation timelines, configuration management procedures

• Produce vendor contribution specifications: what each hardware, software, and platform vendor must deliver to the LLD, and how the technical integration works across vendor boundaries

Design review and approval

• Review and approve architectures produced by Solutions Architects, delivery teams, and GSI or resell partners (Deloitte, Accenture, WWT) before implementation

• Validate that designs conform to Reference Architecture patterns, with particular attention to AI Execution layer standards and Shared Data layer integration

• Approve or reject with written rationale and remediation steps. No design proceeds without TDA sign-off

• Assess whether a design accounts for the partner's actual operational capability, not the capability an elegant design assumes

• For regulated and sovereign environments, validate compliance posture, data residency, and audit evidence design

Technical decisions

• Make binding calls on architecture trade-offs: endpoint selection, resilience levels, intelligence product locality, agent autonomy thresholds

• Resolve disagreements between delivery teams and partner architects while holding 7SG standards

• Decide the infrastructure questions: whether power, space, and cooling are sufficient, which switching fabric fits the workload profile, whether hyperconverged (Nutanix, for example) or disaggregated storage is right

• Decide the data questions: when to build intelligence products from source systems rather than existing BI platforms, latency requirements for data connectivity, vector against graph against structured representation for a given use case

• Issue go and no-go decisions on designs that miss production standards, including when delivery timelines are at risk

Risk identification

• Catch design flaws before implementation money is spent: scalability limits, single points of failure, security gaps, cost structures that break at scale, operational burden the partner cannot sustain

• For each accepted risk, document the trade-off rationale, the agreed tolerance, the monitoring trigger, and the escalation path

• Validate that FinOps attribution is designed into the architecture before deployment, down to the tenant

• Assess the AI-specific risks: model drift detection, intelligence product staleness, grounding failures, agent behavioral anomalies, workload deviation from profiled baselines

Standards, governance, and continuous improvement

• Maintain and evolve design review checklists, approval criteria, and technical guardrails aligned to Reference Architecture principles

• Document the design pattern library: approved architectural approaches and anti-patterns by layer

• Contribute Reference Architecture updates based on what production teaches and what the technology does next

• Work with the Tech Strategy team on architectural direction, industry trends, and vendor alignment (NVIDIA, Databricks, Palantir, )

• Coach delivery teams and Solutions Architects so initial design quality improves and review cycles drop

• Run the design process the way we tell customers to run deployments. Metrics and feedback loops are in place, and the process improves on them

Qualifications

Required experience and expertise

• 12 to 15+ years building production architectures across heterogeneous environments. Hands-on design and implementation, not advisory

• Infrastructure: GPU cluster design (accelerator sizing, switching fabrics, storage subsystems), Kubernetes platform engineering (operators, CSI/CNI, scheduling, multi-tenancy), hyperconverged platforms, hyperscaler IaaS/PaaS (AWS, Azure, GCP), network architecture (data center fabrics, SASE), infrastructure provisioning and hardening, BC/DR across multiple environments

• Data: data lakes, warehouses, and feature stores, vector and graph databases, streaming and micro-batch pipelines, CDC and data connectivity patterns, data governance, lineage, and classification, BI platform integration

• AI/ML and application: model training and inference pipelines, fine-tuning and adapter lifecycle, MLOps, agentic systems (LLM orchestration, tool invocation, RAG, guardrails), model routing and workload placement, API design and integration patterns, ASPM/DSPM, identity and policy enforcement

• Production operations: observability and telemetry design, incident response, FinOps and capacity planning, SLA management, CVE management and hardening

• You have designed for more than one environment. On-prem, hyperscaler, and sovereign estates each behave differently and you have seen at least two of them in production

• Multi-tenant platform design, where someone other than the operator consumes the capacity

• Regulated industry experience (telecom, financial services, healthcare, or government), including compliance, audit, and data residency requirements

• You have held technical authority: blocked bad designs, made unpopular calls, owned the quality outcome

Nice to haves

• Telco or neocloud experience. Useful, and not required. The architecture skill and the production track record matter more than the logo on the badge

• TMF interfaces, proving ground and pilot environments, IaaS/PaaS portal design

• Enterprise architect or principal engineer at a company serving Fortune 500 customers

• GSI partnership experience (Deloitte, Accenture, WWT)

• Creating, maintaining and deploying against validated designs

• Vendor architecture alignment (NVIDIA, Databricks, Palantir, )

• Startup or scale-up experience, so ambiguity and incomplete information are familiar ground

• Working knowledge of AI governance frameworks: inventory, lifecycle management, staged autonomy

About 7sg Inc.

7SG builds and supports the computing infrastructure that runs AI workloads for telcos and neoclouds, licenses the software that moves AI applications from pilot to production and provides the advisory work that tells them how to operate it.

On the infrastructure side that means low-level design of the data centre environment, structured cabling and field certification, oversight of GPU and server installation, the customer portal that provisions capacity, and 24x7 on-site support once the environment is live. The hardware belongs to the customer. We are paid for the engineering, the integration, and the managed service.

On the software side we license a platform that runs on the customer's infrastructure and manages the AI workloads across it: what is running, what it costs, and where each workload should run. The advisory work is assessment and design. We evaluate the environment a customer has, define the target architecture, and produce the roadmap they build against.

The first deployment is with a national telco in the UAE, under a multi-year agreement covering design, build, licence, and field services.