Skip to content

Careers / AI Engineering

AI Engineer

Design, build and ship production AI systems (LLM applications, agents, and real-time perception for robots) for clients in consulting, software and robotics.

Apply now
Open in Gmail
Recruitment processRemote / Hybrid · Full-time · Posted

About the role

As an AI Engineer at ALL FOR AI, you turn frontier models into reliable products. You will design context and prompts, build agents that use tools, set up evaluation, and take systems from prototype to production for our clients.

On robotics projects, you will also build perception and autonomy software in C++ and ROS 2: detecting objects, estimating their 3D pose, and helping robots pick, place and navigate reliably in factories and warehouses.

You will work AI-native every day: pairing with coding agents, automating your own workflow, and sharing what works with the rest of the company.

What you will do

  • Design and build LLM-powered applications, agents and multi-agent workflows end to end.
  • Engineer context: decide what information, tools, memory and instructions a model sees, and in what form.
  • Write, test and version prompts and system instructions as production code.
  • Build retrieval-augmented generation (RAG) pipelines: chunking, embeddings, hybrid search, re-ranking.
  • Create evaluation suites (offline evals, LLM-as-judge, regression tests) and use them to guide every change.
  • Optimise cost, latency and quality through model selection, caching, batching and routing.
  • Implement guardrails, safety checks and human-in-the-loop review where decisions matter.
  • Integrate models with client systems through APIs, tool calling and the Model Context Protocol (MCP).
  • Run client workshops and explain technical trade-offs clearly to non-technical stakeholders.
  • Develop real-time perception for robots in C++ and Python: object detection, segmentation and 6-DoF pose estimation.
  • Build, test and deploy ROS 2 nodes, and integrate cameras, depth sensors and LiDAR with robot controllers.
  • Evaluate state-of-the-art computer-vision research, then implement, harden and benchmark it for production.
  • Profile and optimise algorithms for latency and robustness on systems that run 24/7.
  • Validate perception and planning on real hardware and in simulation, and track performance with automated tests and CI.
  • Contribute prompts, agents and best practices to the company-wide AI library.

Requirements

We expect strong, demonstrable knowledge across all of the areas below.

01Context engineering

  • Deep understanding of how context windows work, including limits, position effects and long-context behaviour.
  • Designing context assembly: system instructions, retrieved documents, tool results, conversation history and memory.
  • Context compression, summarisation and pruning strategies for long-running agents.
  • Structuring information for models: schemas, XML/Markdown delimiters, examples and metadata.
  • Memory design for agents: short-term scratchpads, long-term stores and retrieval policies.

02Prompt engineering

  • Writing clear system prompts, role definitions and task instructions.
  • Zero-shot, few-shot and chain-of-thought prompting, and knowing when each helps.
  • Structured output: JSON schemas, function and tool calling, constrained decoding.
  • Prompt chaining, decomposition, self-critique and reflection patterns.
  • Systematic prompt iteration driven by evaluations, not intuition.
  • Awareness of prompt-injection risks and how to defend against them.

03AI history and foundations

  • The history of AI: Turing's 1950 paper, the 1956 Dartmouth workshop, symbolic AI and expert systems.
  • The perceptron, the AI winters, backpropagation and the revival of neural networks.
  • The deep-learning era: ImageNet and AlexNet (2012), word embeddings, sequence-to-sequence models.
  • The Transformer ('Attention Is All You Need', 2017), BERT, the GPT series and scaling laws.
  • RLHF, instruction tuning, constitutional and preference-based alignment methods.
  • The current landscape of frontier and open-weight models, reasoning models and multimodal systems.

04LLM and machine-learning fundamentals

  • Tokenisation, embeddings, attention, and how decoding and sampling parameters affect output.
  • Fine-tuning, LoRA/PEFT, distillation and when they beat prompting.
  • Classic ML concepts: overfitting, train/validation/test splits, metrics and baselines.
  • Hallucination, bias and failure modes, and how to measure and reduce them.

05Agents, RAG and tooling

  • Building agents with tool use, planning loops and error recovery.
  • Model Context Protocol (MCP) servers and clients.
  • Vector databases and search: pgvector, Pinecone, Weaviate, Elasticsearch or similar.
  • Agent and LLM frameworks, and the judgment to know when not to use one.

06Robotics engineering (C++ and ROS 2)

  • Strong modern C++ (C++17/20): memory management, templates, concurrency and performance profiling.
  • ROS 2: nodes, topics, services, actions, parameters, launch files, lifecycle nodes and QoS settings.
  • Building and testing ROS 2 packages with colcon, ament and CMake; debugging with rviz2, rosbag2 and ros2 CLI tools.
  • TF2 coordinate frames, URDF/xacro robot descriptions and kinematics.
  • Integrating sensors (RGB, stereo and depth cameras, LiDAR, IMU, GNSS) and robot arms or mobile bases.
  • Motion planning and manipulation basics: MoveIt 2, inverse kinematics, collision checking, grasp planning.
  • Real-time and embedded constraints: deterministic latency, multithreading, GPU acceleration (CUDA) and edge devices.
  • Simulation and digital twins: Gazebo, Isaac Sim or similar, including sim-to-real transfer.

07Computer vision and 3D perception

  • Object detection and instance segmentation with modern deep-learning models.
  • 6-DoF pose estimation of textured and textureless objects in cluttered scenes.
  • Camera models, calibration (intrinsic, extrinsic, hand-eye) and multi-view geometry.
  • 3D data processing: point clouds, registration (ICP), meshes and depth maps, using libraries such as OpenCV, PCL and Open3D.
  • Linear algebra, 3D transformations (rotations, quaternions, SE(3)) and statistical analysis.
  • Building datasets, labelling pipelines and synthetic data, and measuring accuracy with clear metrics.
  • Deploying vision models efficiently: ONNX, TensorRT, quantisation and batching.

08Software engineering

  • BS, MS or PhD in computer science, robotics, electrical engineering or a related field, or equivalent experience.
  • 3+ years of professional software engineering, strong in Python and C++ (TypeScript is a plus).
  • Designing APIs, background workers and event-driven systems.
  • Testing, CI/CD, Git workflows and code review.
  • Cloud deployment (AWS, GCP, Azure or Vercel), containers and observability.
  • Security fundamentals: secrets handling, data privacy and access control.

09AI-native working style

  • You already use AI assistants and coding agents daily, and can show workflows you built.
  • You verify AI output rigorously and can explain every line you ship.
  • Clear written and spoken communication in English.
  • Curiosity: you follow new models and research and test them quickly.

Nice to have

  • Experience with multimodal models (vision, speech, documents).
  • Experience with SLAM, visual odometry or object tracking.
  • Robotic picking, palletising or depalletising in logistics or factory automation.
  • Computer graphics, rendering or augmented reality.
  • Research publications in robotics or computer vision (CVPR, ICCV, ECCV, ICRA, IROS, RSS).
  • Consulting or client-facing delivery experience.
  • Open-source contributions or published writing on AI engineering.
  • Experience running open-weight models (vLLM, quantisation, GPU serving).

Recruitment process

  1. 1

    Apply

    Send your CV and a short note to careers@allforaicompany.business. Writing it with AI is welcome.

  2. 2

    Screening call

    A 30-minute call with our talent team about your experience, motivation and the AI tools you use.

  3. 3

    One AI-assisted interview

    A single 90-minute interview, structured as a simulated client engagement. Use any AI tool you want.

  4. 4

    Decision & offer

    We share feedback and a decision within five business days of the interview.

The interview: a simulated client engagement

Our process has one interview. It simulates a real ALL FOR AI client engagement, so you show us how you would work on the job. You may use any AI assistant, model, agent or tool throughout, in line with our Interview AI Assistance Policy.

A few days before the interview, you receive a briefing email from careers@allforaicompany.business. It describes a fictional client, its business problem and what the client expects from you. Scenarios are modeled on the kinds of clients we serve.

Scenarios, agenda and assessment