AI Engineer/ML Engineer - Senior Developers - AI Training - Glasgow, UK

prolificacademicltd; United States; Remote; USD 55,000 - 110,000; Part time

Benefits offered by this job

Competitive hourly rates

Remote-first, flexible schedule

Lightweight onboarding

Job summary

Prolificacademicltd is seeking experienced AI/ML engineers to contribute to rigorous evaluation of large language models. Tasks are paid and designed around roughly one hour of focused work, with a fully remote, asynchronous workflow that fits a practicing engineer's schedule.

You will review model explanations, examine production ML code, provide high-quality human feedback for RLHF, and help benchmark outputs against defined taxonomies and metrics. Remote, flexible, and task-based.

Qualifications

• Degree in a quantitative discipline with ML emphasis.
• Experience building or fine-tuning ML models in production.
• Deep expertise in neural architectures and optimization methods.
• Experience with prompt engineering or RLHF workflows.

Responsibilities

• Evaluate explanations of architectures, loss functions, and backpropagation for technical soundness.
• Inspect ML code and notebooks, including training pipelines and evaluation scripts, for efficiency and correctness.
• Provide high-quality human feedback for RLHF to improve model alignment with human intent.
• Trace reasoning and flag points where model logic fails or hallucinations occur.
• Design and execute benchmark tests across model outputs using defined metrics.

Skills

ML production experience

Neural networks architectures

Prompt engineering / RLHF

Code review & debugging

Education

BS/MS/PhD in CS/ML or related field

Tools

PyTorch

TensorFlow/Keras

Hugging Face Transformers

AWS SageMaker

Job description

Role overview

An expert contributor network is recruiting experienced AI and Machine Learning Engineers to help train and rigorously evaluate large language models. Contributors receive paid technical tasks that leverage real-world ML production experience, with most sessions designed to fit within an hour of focused work. The arrangement is fully remote, asynchronous, and built to accommodate practicing engineers.

Responsibilities

• Evaluate AI-produced explanations of model architectures, loss functions, and backpropagation for technical soundness
• Inspect ML code and notebooks, including training pipelines, preprocessing logic, and evaluation scripts, for efficiency and correctness
• Provide high-quality human feedback used in RLHF training to improve model alignment with human intent
• Trace chain-of-thought reasoning and flag specific points where model logic collapses
• Design and execute comparative benchmark tests across model outputs against defined taxonomies and metrics

Requirements

• Degree (BS, MS, or PhD) in Computer Science, Artificial Intelligence, Robotics, or another quantitative discipline with a machine learning emphasis
• Demonstrated experience building, deploying, or fine-tuning ML models in production settings
• Deep expertise in neural network architectures (Transformers, CNNs, RNNs) and contemporary optimization methods
• Practical experience with prompt engineering, RLHF, or retrieval-augmented generation workflows
• Strong technical rigor when auditing model logic, spotting training data contamination, and evaluating mathematical foundations of algorithms
• Meticulous attention to detail for identifying hallucinations, biased outputs, or reasoning failures

Nice to have

• Expert-level command of PyTorch or TensorFlow/Keras
• Advanced Python fluency with NumPy, Pandas, and Scikit-learn, plus Hugging Face Transformers experience
• Cloud and MLOps exposure with AWS SageMaker, Google Cloud Vertex AI, Weights & Biases, or LangChain
• Familiarity with vector stores such as Pinecone, Milvus, or Weaviate for evaluating RAG pipelines

Benefits and work setup

• Competitive hourly rates, with researcher-driven tasks typically paying up to $80 per hour
• Remote-first, flexible schedule; contributors log in and pick up tasks as available
• Lightweight onboarding: a 10-15 minute assessment unlocks the contributor pool, with most tasks capped at roughly one hour of uninterrupted work