AI Engineer

MAS Global Consulting; Colombia; On-site; COP 90,000,000 - 120,000,000; Full time

Job summary

MAS Global Consulting, LLC is seeking engineers who move fast yet understand the problem, user, and client. You will build MAS's AI platforms end to end, from frontend and API to data and deployment, sitting with stakeholders to define what to build and staying close throughout development.

The ideal candidate has solid full-stack experience with TypeScript and Python, ships products, and can explain decisions and tradeoffs.

Qualifications

• Solid full-stack experience with shipped products.
• Proficiency in TypeScript and Python on production systems.
• Experience building AI-enabled platforms and agentic workflows.
• Strong stakeholder communication and ability to translate needs into technical solutions.
• Ability to own features end-to-end from planning to deployment.

Responsibilities

• Build MAS AI platforms end to end: frontend, API, data, deployment.
• Engage with stakeholders to understand problems before building.
• Ensure high-quality user experience and maintainable code with agentic SDLC.
• Automate processes and measure improvements.
• Create prototypes for demos and pilots when needed.

Skills

Full-stack development

TypeScript

Python

Shipped products

Stakeholder communication

English communication

Tools

Azure

Job description

Software delivery has changed. At MAS we build with AI agents as part of the team, and with people always making the calls that matter. We are looking for the engineers who will take MAS to the next level: engineers who are fast because they understand the problem, the user and the client, not instead of understanding them.

What you will do

• Build MAS's AI platforms and products end to end: frontend, API, data, deployment.
• Sit with stakeholders and users to understand the problem before deciding what to build, and keep them close while you build it.
• Sweat the experience: what you ship has to work well and feel well made, considering best agentic SDLC practices
• Automate processes where it clearly pays off, and measure the before and after.
• Build prototypes for demos and pilots when the moment calls for it, and keep the line between prototype and product explicit.

How we work

• As a team. Engineers, designers, business people and AI agents build together; nobody ships alone and nobody hides behind a ticket.
• Agentic, with a human in the loop, always. Agents plan, implement and review with us, people define the intent and hold the judgment.
• “The AI wrote it that way” does not explain a bug or a missed requirement. If you shipped it, you own it. Agent output is a draft you judge before it ships.
• We ship small and early: a working version in front of someone this week beats a full plan next month.

What we look for

Speed only counts if you understand what the client needs, what a good experience looks like, and why the problem matters.

Must have

• Solid full-stack experience, TypeScript and Python in our case, with shipped products you can walk us through: the decisions, the tradeoffs, what you would do differently.
• Advanced agent orchestration: you have designed and shipped multi-step or multi-agent workflows (planning, tool use, handoffs, retries, human checkpoints) and can explain why that shape beat a simpler one.
• An agentic SDLC in daily practice: spec, implementation, tests and review with agents in the loop, under gates that catch what the agents get wrong. We will ask you to show us how you work.
• Stakeholder fluency: you can sit with a client or a business team, understand the problem in their words, and translate it into what to build — and what not to.
• Ownership past the pull request: you follow whether what you shipped got adopted, and you change course based on what you learn.
• Working English for stakeholder conversations, code review and documentation.

Good to have

• Advanced RAG: retrieval pipelines you built and tuned (chunking, indexing, reranking) and the measurements that told you they improved.
• Evaluation practice: eval sets, LLM-as-judge or regression harnesses that gated releases.
• Cloud depth, Azure in our case: infrastructure as code, observability, cost-aware deployment of AI workloads.