Principal AI Architect

We are looking for a Principal AI Architect to lead architecture across MathCo's enterprise AI engagements. You will set the architecture standard for a portfolio of client programs, act as the senior technical counterpart to client technology leadership, and stay hands-on, building with AI codingagents and reference implementations rather than only reviewing others' work. The role combines architecture authority, technical solutioning for new pursuits, and the development of the architects and forward-deployed engineers who deliver it

Key responsibilities

Architecture Authority Across the Portfolio: Own the target architecture for a portfolio of enterprise AI engagements. Set reference architectures and patterns for retrieval, agentic and decision-intelligence systems on Gemini /Google Cloud, Databricks and Azure AI, and decide where platform-native capability is sufficient and where custom build is justified.

Chair architecture reviews across engagements. Approve solution designs and ADRs from AI Architects and FDE teams, and stop designs that will not hold up in production on accuracy, latency, cost, security or failure-mode grounds.

Personally design the hardest systems: multi-agent orchestration, LLM-inclusive systems integrated with enterprise estates, and evaluation and guardrail layers, especially where the problem is novel or the cost of error is high.

Client Technical Leadership & Solutioning

Act as the senior technical counterpart to client CxOs, VPs and enterprise architects. Frame AI architecture choices in business terms (platform vs. custom, accuracy vs. cost, scope vs. timeline) and hold a defensible position under scrutiny.

Lead technical solutioning on pursuits: shape the architecture, delivery approach, effort and risk profile behind proposals, so that commercial commitments, including compressed timelines, rest on an architecture that can meet them.

Assess client data and AI estates (platform readiness, data and access, governance) at engagement start, and set the path from pilot to production at scale.

Hands-On, AI-Accelerated Engineering: Remain a practicing engineer. Build reference implementations, spikes and proof points with AI coding agents, and use them to set the standard expected of delivery teams.

Define how teams use AI coding agents across the SDLC (specification, parallel build tracks, review, hardening and testing) so that throughput gains arrive at a production quality bar, not a prototype one.

Evaluate new models, agent frameworks and platform releases on Gemini, Databricks and Azure AI through hands-on testing, and turn the findings into guidance teams can adopt.

Production, Governance & Security of AI Systems: Set the production bar for AI systems across engagements: evaluation harnesses and thresholds, versioned prompts and models, CI/CD, observability, rollback and cost controls (LLMOps /AgentOps).

Define the security and responsible-AI architecture for agentic systems: agent identity, scoped short-lived credentials, human approval for high-risk actions, auditable traces, data privacy and regulatory controls.

Own architecture-level risk on engagements. Escalate early when design, data or platform constraints threaten committed outcomes, and present options as priced trade-offs.

Accelerators, AI Tooling & Platform Partnerships: Watch industry practice and recurring patterns across client engagements, and turn them into MathCo accelerators and internal AI tools: reference implementations, agent templates, evaluation kits and coding-agent skills. Build them to production standard and maintain them as products, not one-off assets.

Spot where an accelerator fits an engagement or pursuit and push for its adoption. Make the case to clients on trade-offs and benefits (time to value, cost, risk, customization limits and dependency) and to delivery teams on fit and integration effort.

Serve as MathCo's architecture lead with at least one strategic platform partner (Google Cloud / Gemini, Databricks or Microsoft Azure AI). Track platform roadmaps, shape joint solutions, and keep MathCo's accelerators aligned with where the platform is heading.

Maintain the reference architecture library, standards and failure register, and make sure each engagement returns assets and lessons so the next one starts further ahead.

Talent & Practice Leadership: Develop L5 AI Architects and FDEs through design reviews, pairing and structured feedback, and build an architecture bench that can run engagements without escalation.

Contribute to the architecture practice's hiring bar, certification paths, and point-of-view content for clients and the market.

Required qualifications & experience

16-18 years of total experience across data science, data engineering, software engineering and AI.

At least 10-12 years of hands-on work earlier in your career as a data scientist, ML engineer, data engineer or software engineer: building models, pipelines, data platforms and applications yourself and taking them to production, not only overseeing them.

5+ years in technology leadership: leading data and AI programs across multiple teams, owning technical outcomes, and leading senior data scientists, data engineers and architects.

Still hands-on, particularly with AI tools. You build with AI coding agents, LLM APIs and agent frameworks as part of your regular work, and can show recent working systems.

The last 4 years spent on AI projects, including implementing enterprise AI solutions (ML-driven decision systems, retrieval and agentic applications) through to production, integrated with enterprise data, security and identity estates.

Very good familiarity with at least one of: Google Cloud / Gemini (Vertex AI, Gemini models), Databricks (Mosaic AI, Unity Catalog) or Microsoft Azure AI (Azure AI Foundry, Azure OpenAI), backed by projects delivered on that platform.

Strong grounding in data platform architecture: lakehouse and warehouse design, distributed processing on Spark, data modeling, semantic layers, data quality and governance.

Demonstrated experience designing AI systems in which LLMs are one component among many (retrieval and context pipelines, agent orchestration, evaluation harnesses, guardrails), as distinct from building solely with managed toolkits.

Working knowledge of MLOps, LLMOps and AgentOps, security for AI systems, and responsible[1]AI and regulatory requirements (for example SOC 2, HIPAA, GDPR, EU AI Act).

A track record as the senior technical counterpart to client CxOs and VPs, including technical solutioning on pursuits; consulting or professional-services experience preferred.

Professional-level certification on at least one MathCo-preferred platform (Google Cloud, Databricks, Azure, AWS or Snowflake) preferred.