AI architect and tech lead: the senior track at a GCC
The AI architect role at a GCC explained: scope, system design, build versus buy, stakeholder work, the skills expected and how senior engineers move into it.
On this page
An AI architect or tech lead designs end-to-end AI systems and guides teams on cost, risk and strategy. The AI architect role is the senior track at a Global Capability Centre (GCC): the work shifts from building one component well to deciding what should be built, how the parts fit together and whether the result justifies its cost. The role usually goes to engineers with around ten years or more of experience, including real time spent building AI systems.
The architect track is one of seven in the overview of AI engineering roles at GCCs.
What does an AI architect or tech lead do?
An AI architect or tech lead makes the decisions that are expensive to reverse, and makes sure the team understands them. The output is designs, reviews, standards and judgement, and only occasionally code.
A typical week might include:
- Writing a design document for a document assistant that must serve several business units with different access rules.
- Reviewing a team's plan to fine-tune a model, and asking whether retrieval would solve the problem at lower cost.
- Comparing a vendor product with an in-house build for a contact-centre use case, and presenting the trade-offs to leadership.
- Agreeing with risk and compliance colleagues what data a model may see and what must be logged.
- Estimating the running cost of a feature at full scale before the team commits to it.
- Unblocking two teams that have built incompatible retrieval pipelines, and setting a shared pattern.
A GCC adds a particular demand: the stakeholders are often in the parent company's headquarters, in another time zone, with their own priorities. An architect in India is expected to own the design and argue for it, not to wait for one to arrive. How a GCC AI team works explains that relationship.
What skills does an AI architect need?
An AI architect needs breadth across the whole AI stack, plus a group of leadership skills that no other track treats as core. Depth in every topic is neither possible nor expected. Knowing enough about each to make a sound decision is.
| Skill area | What the architect track treats as core |
|---|---|
| Engineering | Python, APIs and web services, system design, distributed systems |
| Data | Data governance and privacy, lakehouse and the modern data stack |
| Models | ML fundamentals, model evaluation, LLM fundamentals, LLM evaluation |
| LLM applications | RAG, agents and tool use, multi-agent systems, knowledge graphs, AI safety and security |
| Platform | Cloud platforms, cloud AI platforms, cost and performance |
| Design | ML system design, GenAI architecture |
| Leadership | Problem framing, build versus buy, responsible AI, technical communication, technical leadership |
Notice what sits at a lower level for the architect: PyTorch, hands-on feature engineering and day-to-day Kubernetes. An architect should understand them and need not practise them daily. The maths of machine learning is good to know at this level, not core.
For the design skills, a framework for the ML system design interview is the best starting point, and what RAG is and how it works and how to evaluate an LLM application cover the two LLM topics architects are asked about most.
How does an AI architect decide between build and buy?
An AI architect decides between build and buy by asking whether the capability sets the company apart, and whether the company can afford to own it. Build versus buy is the most frequent strategic question in the role, and the honest answer is often a mixture.
Lean towards buying or using a managed service
- The capability is common across companies, such as transcription or generic search
- The team is small and cannot run it at night
- Speed to a first release matters more than control
- The vendor meets the company's data and compliance rules
Lean towards building in-house
- The capability depends on proprietary data or domain knowledge
- Data cannot leave the company's environment
- Usage is high enough that per-request pricing becomes the main cost
- The team needs to change behaviour faster than a vendor allows
Whatever the decision, a good architect writes down the exit: what it would take to change course in a year. A design that isolates the model behind a clear interface keeps that option open.
How is the architect role different from a senior engineer role?
The architect role differs from a senior engineer role in the unit of work: a senior engineer is accountable for a system, and an architect is accountable for a set of decisions across systems. A senior ML or GenAI engineer can succeed by building excellent things. An architect succeeds only if other people build the right things.
Three shifts follow from that difference:
- From answers to trade-offs. A senior engineer finds the best solution. An architect explains why a second-best solution is the right one given cost, risk and time.
- From code to writing. Design documents, decision records and review comments become the main output.
- From one team to many audiences. The same design must be explained to engineers, to a product owner and to a risk officer, each in their own terms.
Engineers who want seniority without these shifts can usually stay on a specialist path, growing as a senior or staff ML engineer or GenAI engineer.
Is the AI architect role right for you?
The AI architect role is right for you if you enjoy deciding and persuading as much as building. Seniority alone does not make the role a good fit.
The role tends to suit:
- Senior ML engineers and platform engineers who already get pulled into design reviews across teams.
- Architects and tech leads from service companies who have delivered large systems and have since built AI systems hands-on.
- People who can stay patient through governance, security and budget discussions.
The role tends not to suit:
- Engineers who want to code most of the day.
- Experienced architects with no hands-on AI work. Designs drawn from slides fall apart in a system design round.
- People who dislike being accountable for outcomes they do not directly control.
How do you move into an AI architect role?
You move into an AI architect role by earning the breadth first-hand and then showing the judgement. Long experience gets your profile read; the system design and leadership rounds decide the offer.
Build across the stack
Ship at least one conventional model and one retrieval system yourself, including evaluation and monitoring.
Learn the costs
Estimate and then measure what a system costs to run. Architects are asked about money early.
Write design documents
Record the options, the choice and the reasons for real decisions, at work or on your own projects.
Practise governance
Learn how privacy, access control and responsible AI reviews change a design.
Lead without the title
Run design reviews, mentor engineers and settle a cross-team disagreement.
Senior hiring usually includes one or more system design rounds and a long conversation with a hiring manager about past decisions. The hiring manager round explains what is being judged there, and how to move from a service company to a GCC covers the move for senior service-company engineers.
You can see the architect track beside the other six in the role explorer, or browse which centres exist in your field in the GCC directory.
Key takeaways
- An AI architect is accountable for decisions and designs, not for individual models or services.
- Breadth across data, models, LLM applications and infrastructure matters more than depth in one.
- Build versus buy, responsible AI, problem framing and technical leadership are core skills for this track only.
- The role is a poor fit for people who want to stay hands-on most of the day.
- Existing architects need hands-on AI experience before the title is credible.
Frequently asked questions
Can a solution architect become an AI architect?
Yes, and a solution architect brings much of what the role needs: system design, cloud platforms, stakeholder work and cost thinking. The missing part is first-hand AI experience. Build and evaluate at least one retrieval system and one conventional model yourself, so that your designs rest on how these systems really fail and not on vendor diagrams.
Does an AI architect still write code?
An AI architect writes less production code than an engineer, but should still write some: prototypes, proofs of concept and reviews of critical code. Python is treated as core for the track. Architects who stop building tend to lose touch with what is easy and what is hard, and their estimates and designs suffer.
Is AI architect a management role?
Not usually. An AI architect or tech lead leads through design and influence, often without direct reports, while a manager is responsible for people, hiring and performance. Some GCCs combine the two in one lead role. Ask in the interview how much of the job is people management and how much is technical direction.
What is the difference between an AI architect and a tech lead?
A tech lead usually guides one team and stays close to its code and delivery. An AI architect usually works across several teams or a whole platform, setting patterns and standards. The site's track groups them because both are judged on design decisions, trade-offs and guiding other engineers.
Do AI architects need certifications?
Cloud certifications can help show platform knowledge, but they do not make an AI architect. Hiring panels for senior roles usually look for systems you designed, decisions you made under constraints and what happened afterwards. A clear account of one difficult design choice carries more weight than a list of certificates.