Chi tiết công việc AI Engineer tại Mileon
AI Engineer
mileon
Hanoi, Vietnam · Full-time · Office-based · 15–30M VND/month
The problem we’re solving
Mid-market and enterprise businesses are drowning in tools they barely use. They’ve bought CRMs, project management platforms, analytics dashboards, and automation software — but none of it talks to each other, and none of it taps into what AI can actually do. The result is shallow usage across the board: teams manually copy data between systems, write proposals from scratch every time, chase leads that go cold because nobody followed up, and spend hours on tasks that should take minutes. The tools exist. The integration doesn’t. The intelligence doesn’t.
Mileon builds, runs, and maintains the AI systems that fix this. Not chatbots. Not wrappers on top of ChatGPT. Full production systems that replace manual processes businesses used to staff entire teams for. We take a company’s messy or complex operations, identify where AI creates real leverage, and build integrated systems that run continuously — automated lead follow-up, proposal generation, CRM management, operational reporting, and more.
About Mileon
Mileon is an AI systems integrator. We work with mid-market and enterprise clients across the Asia Pacific region to build AI-powered operational systems that run in production, not demos that look good in a pitch.
We don’t hand clients a tool and walk away. We build the system, we operate it, the client gets outcomes — leads followed up, proposals generated, operations running without manual intervention. Each module we deploy feeds into the next, creating compounding value: the data from one layer makes the next layer smarter, and the whole system becomes something the client couldn’t replicate or replace with off-the-shelf software.
We’re a growing team, three months old as a standalone company but spun off from an established software services firm. You’d work directly with our lead engineer and co-founder. The engineering team is small and senior — no layers between you and the people making decisions. If your idea is good, it ships this week.
What you’d build
The work spans whatever the client needs automated. Some examples from real engagements:
Automated assessment pipelines — a prospect submits a form and within minutes gets a full enrichment report: industry analysis, workflow mapping, competitor landscape, website audit, and a redesign proposal. End-to-end, no human in the loop.
Operational systems for service businesses — automated lead follow-up, proposal generation from templates and client data, CRM management, invoice automation, and operational reporting. The systems that close the gap between leads delivered and revenue captured.
E-commerce automation — Shopify migration systems, catalog operations, auto-generated product content, and merchandising automation. Built once, runs continuously across the store.
Agent orchestration — multi-step LLM workflows that coordinate tools, pull from APIs, and produce structured deliverables. The kind that run unattended and handle real client data reliably.
Tooling evaluation and infrastructure — we’ve tested multiple model providers, orchestration layers, and harness architectures. You’d run real benchmarks (cost, quality, reliability), make the call on what we standardise on, and find cost optimisations like the ones that have already cut our API spend significantly.
Your first 90 days
Day 1–30
Get into the stack, the clients, and the current systems. Ship one end-to-end automation pipeline to production. Form an opinion on what to keep, what to replace, and where the bottlenecks are.
Day 31–60
Own at least one client’s AI system end-to-end. Run your first tooling benchmark with a written recommendation. Start reducing turnaround time on deliverables from days to hours.
Day 61–90
Be the person the team trusts to make AI architecture decisions and solutions independently. You’re building and shipping without supervision, and the systems you own are running reliably in production.
The right team member vs. the wrong one
The wrong team member takes a requirement like “generate a customer profile from this form” and prompts an LLM, copies the output, and calls it done.
The right team member asks: Why is this manual? What if the form submission triggered an enrichment pipeline that also pulled industry benchmarks, mapped the client’s workflow, identified automation opportunities, and delivered a proposal — all before anyone touches it? And then they build it.
The difference isn’t intelligence. It’s that the right person treats the output standard as their own reputation. They think in systems, not tasks. They find the 10x lever without being told to look for it. They understand that the real product isn’t one automation — it’s the integrated system that compounds over time.
What you must bring
Hands-on LLM building experience — Claude, GPT-4, or equivalent. Not just prompting: actual API integration, tool use, function calling, agent orchestration. You’ve deployed something that runs without you watching.
Production software engineering — Python and/or TypeScript. Backends, not notebooks. You can build, deploy, and maintain systems that handle real data.
Systems thinking — when someone asks for a one-off, you ask why it isn’t automated. When a pipeline breaks, you fix the root cause. When you see three manual steps, you see one system.
Cost-aware engineering — you think about token costs, model selection, and infrastructure spend as design constraints, not afterthoughts.
Independent judgment — you don’t wait for someone to tell you which tool to use. You research, evaluate, form an opinion, and back it with evidence.
Nice to have
Experience with Anthropic’s ecosystem (Claude Code, MCP servers, Agent SDK)
Background in e-commerce platforms (Shopify, WooCommerce) or website migration
Exposure to CRM/enrichment tools (HubSpot, Apollo, Clay)
English fluency — strong bonus. We operate across the Asia Pacific region and English is our working language
How we work
Mileon is a growing team, three months old as a standalone company but spun off from an established software services firm. We exist to do one thing: implement AI systems that increase quality of life on a global level. We believe in the mission. We move fast. We radically address bottlenecks — even when the bottleneck is you. We are output-driven.
No sprint ceremonies for the sake of it. No ticket queue with 400 items. You talk to the founders directly. If your idea is good for the client, it ships this week. You’ll have deploy access and commit access to every repo from day one.
We don’t do performance reviews. We do quarterly conversations about what’s working and what’s not. Training budget is open — propose what you want to learn and how it helps, and we’ll fund it.
We measure output, not hours. The office is home base, but we serve clients across time zones — starting with Australia. There’s no 9-to-5 rigidity. If a deliverable is due, you get it done. If a client needs a call outside standard hours, you take it. Travel may be required as the business grows.
Where this goes
You’re joining early at a company that’s growing fast. In 12–24 months, this role evolves based on what you’re best at: technical architect owning the entire systems layer, team lead building and managing the engineering function, or deep specialist in AI infrastructure and tooling. The trajectory is yours to shape — we promote based on output, not tenure.
Details
Location
Hanoi, Vietnam (office-based). Clients are across APAC — flexibility with hours and willingness to travel required.
Type
Full-time independent contractor
Compensation
15–30M VND/month (paid in VND or USDT/USDC, your choice)
Leave
12 days paid leave + Vietnamese public holidays
Perks
Open training budget (proposal-based), discretionary quarterly bonuses
Reports to
Lead Engineer and Co-founder
Start
Immediately
How to apply
Send your CV and a note about what you’ve built with AI to the email below. We’ll review within 48 hours. If there’s a fit, you’ll have a 30-minute technical conversation with the lead engineer, followed by a short paid trial project. No algorithm tests. No panel interviews. We want to see how you think and build.
