AI Tech Lead (Vertex AI and Azure AI Foundry)

Limestone Digital; Brasil; Remoto; BRL 260.000 - 460.000; Tempo pieno

Descrizione del lavoro

Limestone Digital is seeking an AI Tech Lead to own the technical direction for AI agent engagements on US investment firm projects. You will architect agent systems on Vertex AI and Azure Foundry/OpenAI, review code, and guide engineers while coordinating with MSPs and client stakeholders.

The role requires deep cloud and data security expertise, the ability to run discovery sessions, and hands-on contribution during critical deadlines.

Competenze

• 7+ years of software engineering experience with Python.
• 2+ years in tech lead or architect roles.
• Shipped LLM agents or LLM features to production (tool calling, structured output, evaluation harnesses, golden sets, guardrails, cost control).
• Deep hands-on experience with Google Cloud (Vertex AI) or Azure (Foundry/OpenAI).
• Experience with integration architecture (event-driven flows, on-prem SQL read-only access, secrets management, least privilege, observability).
• Infrastructure as code (Terraform) and CI/CD with quality gates.
• Ability to run client technical sessions and explain trade-offs to executives.
• Mentor engineers, set coding and evaluation standards.
• Fluent English (spoken and written).
• Experience with regulated or confidential data (finance, HIPAA).

Mansioni

• Own technical direction and delivery quality on the US investment firm project.
• Pair with and back up the FDE; contribute hands-on when deadlines require it.
• Run technical sessions with client stakeholders and MSP.
• Maintain agent registry and evaluation runs after model updates; keep runbooks current.
• Lead pre-engagement architecture sessions for healthcare projects when applicable.
• Provide estimates and architecture input for new proposals on the same stacks.
• Enforce confidentiality and separate data, code and environments between engagements.

Conoscenze

Python

Technical leadership

LLM agents

Cloud platforms

Terraform

CI/CD

Vertex AI

Azure OpenAI

Observability

Security and confidentiality

Strumenti

Terraform

CI/CD

Vertex AI

Azure Foundry/OpenAI

Descrizione del lavoro

AI Tech Lead (Vertex AI and Azure AI Foundry)

Requirements

• 7+ years of software engineering, strong Python, at least 2 years as tech lead or architect
• Shipped LLM agents or LLM features to production: tool calling, structured output, evaluation harnesses and golden sets, guardrails, cost control
• Deep hands-on experience with Google Cloud (Vertex AI, ADK preferred) or Azure (Microsoft Foundry, Azure OpenAI), and working knowledge of the other
• Designs deterministic pipelines first and puts the model behind narrow, testable jobs
• Integration architecture: event-driven flows, read-only access to on-premise SQL, secrets management, least privilege, observability
• Infrastructure as code (Terraform) and CI/CD with quality gates
• Runs client technical sessions alone, including with client IT or MSP, and explains trade-offs to non-technical executives
• Reviews and mentors engineers, sets coding and evaluation standards
• Fluent spoken and written English
• Experience with regulated or confidential data (finance, HIPAA) and strict confidentiality rules

Nice-to-Have

• Google Workspace APIs, Apps Script, Slack apps
• Microsoft Foundry content filters and prompt shields, Azure networking (private link, VPN, ExpressRoute)
• Recommendation or forecasting logic on transactional data (order cadence, quantities, unit of measure)
• MCP, A2A, model routing across Gemini, Claude and OpenAI models
• Background in private equity, finance operations or healthcare supply chain
• Pre-sales: architecture sessions, estimates, proposal input

Responsibilities

• Own technical direction and delivery quality on the US investment firm project: architecture of each agent, code review, evaluation standards, release decisions
• Pair with and back up the FDE, build hands-on where a deadline needs it
• Run technical sessions with client stakeholders and their MSP
• Keep the agent registry, evaluation runs after model updates, and runbooks current
• US healthcare: run the pre-engagement architecture sessions, lead Part A discovery (data map, golden set, working slice), set the Azure foundation architecture and direct the Senior AI Engineer through the build
• Give estimates and architecture input for new proposals on the same stacks
• Follow each client's confidentiality rules and keep data, code and environments separate between engagements

About the Project

Lead engineer for AI agent engagements in the Velocity Pod model, starting with two US clients on two clouds.

Primary: US investment firm. The client runs on Google Workspace and Slack. We build agents and automations for the executive office and firm operations on Vertex AI: multi-party meeting scheduling with a calendar change log, a new OKR system on Google Sheets, then email triage, travel rebooking, Slack intake of IT requests and monitoring of the client's existing bots. The pod already has a full-time Forward Deployed AI Engineer (separate hiring request). This role owns technical direction and delivery quality, reviews and pairs with the FDE, and builds hands-on where the deadline needs it.

Secondary, if the deal closes: US healthcare, US medical supplies distributor, HIPAA context. Proposal v1.0 sent in September 2026: AI Foundation plus an Easy Order recommendation agent on Azure with Microsoft Foundry and OpenAI models, reading an on-premise SQL reporting database through a read-only query service, acceptance against a golden set, about 2 weeks of discovery and about 12 weeks of build, with three more agents planned on the same foundation. Here this role is the AI Lead named in the proposal: runs the pre-engagement architecture sessions with the client and their MSP, sets the architecture and directs a full-time Senior AI Engineer.

Both engagements follow one principle: compute in code, give the model only narrow jobs (explanation, judgement), and measure every change against a golden set and an evaluation harness.