AI Engineer

Augmentus; Singapore; On-site; SGD 120,000 - 180,000; Full time

Job summary

Augmentus is seeking an experienced ML engineer to design and deploy agent tools and prompts. You will build multi-model orchestration to balance accuracy, latency, and token cost, and own context engineering across prompts and retrieval.

You will fine-tune open-weight models, distill teacher models, and deploy on GPUs for on-premises setups using Docker and PyTorch.

Qualifications

• 2–5 years building ML/LLM systems in production.
• Strong Python and solid engineering practices (testing, Git, APIs).
• Hands-on agent development: tool calling and structured outputs.
• Fine-tuning open-weight models (PyTorch, Hugging Face, PEFT).
• Experience designing evaluations that measure model/agent quality.

Responsibilities

• Design and build agent tools and process skills for planning and execution.
• Orchestrate multiple models for balance of accuracy, latency and cost.
• Own context engineering: prompts, tool schemas, retrieval, and logs.
• Fine-tune and distill models for on-premise deployments.
• Develop evaluation harnesses and release workflows.

Skills

Python

Engineering practices

Agent development

Prompt and context engineering

Education

Computer Science degree (welcome)

Tools

Docker

PyTorch

Hugging Face

Kubernetes

Job description

You'll build the agent and the models behind it, from tool and skill design to fine-tuning and serving open-weight models on customer premises.

Design and build agent tools and process skills that let the agent plan scans, generate robot programs and resolve motion errors. Build multi-model orchestration, routing tasks across large and small models to balance accuracy, latency and token cost. Own context engineering: prompts, tool schemas, and retrieval over process knowledge.

Fine-tune open-weight large language models and vision-language models (supervised fine-tuning, Low-Rank Adaptation and its quantised variant) on domain data. Distill large teacher models into small, deployable models. Serve and optimise models on graphics processing units for on-premises and airgapped deployments (quantisation, batching, vLLM or similar).

Evaluation and reliability Build deterministic evaluation harnesses and benchmark suites; track regressions across model and agent versions. Own model and dataset versioning, experiment tracking and release workflows. Build logging and tracing for agent runs to support debugging and reliability. Curate training and evaluation datasets for agent and model development.

Requirements

2–5 years building machine learning or large language model systems that reached production.

Strong Python and solid engineering practice (testing, Git, code review, application programming interfaces).

Hands-on agent development: tool and function calling, structured outputs, prompt and context engineering.

Hands-on fine-tuning of open-weight models (PyTorch, Hugging Face, parameter-efficient fine-tuning libraries).

Experience designing evaluations that measure model or agent quality objectively.

Experience deploying models (Docker, graphics processing unit inference, cloud platforms).

Good to have

Multi-agent or subagent architectures.

Distillation, quantisation, or on-premises and airgapped deployment.

Vision-language models, 3D point clouds, or computer vision.

Robotics, Robot Operating System, motion planning or industrial automation.

Agile development with Jira.

A degree in computer science or a related field is welcome but not required. We care about what you've built.

What success looks like

3 months: shipped a new agent tool or skill to production, with an evaluation covering it.

6 months: own a fine-tuned or distilled model in the evaluation and release pipeline.

12 months: a self-hosted model you built is running in a customer deployment.