About Always Further

Always Further is an exciting new startup championing Small Language Models (SLMs) as the future of AI. We build open-source tools and infrastructure to make it easy for developers and organizations to create, fine-tune, and deploy specialized AI models that are efficient, transparent, and privacy-respecting. We believe AI should not require a datacenter and it's own energy grid to be useful. We're empowering developers and enterprises to build specialized AI models that deliver precise, reliable results without the cost and inefficiency of oversized general-purpose models. Always Further is backed by leading investors in AI and technology, and our team has built open source projects used by thousands of developers worldwide.

Why join us

• Work on cutting-edge AI technology at the forefront of open source artificial intelligence
• Build in the open - contribute to open source projects that bring value to the AI community
• Join a team with proven track record of building systems used by industry leaders
• Startup environment with the opportunity to make significant impact and own major technical initiatives
• Collaborate with a passionate community of developers, researchers, and AI practitioners
• Available as a part-time role (15–20 hours/week) for PhD students

About the role

We're seeking a Research Engineer to explore, evaluate, and operationalize new research ideas in dataset generation, supervision strategies, and small language model (SLM) fine-tuning. This role is ideal for a current Masters or PhD student who wants applied research experience and the opportunity to work hands-on with real-world data and models.

You will investigate cutting-edge techniques, prototype research concepts, and help translate findings into practical tooling that strengthens our dataset generation and fine-tuning stack, including contributions to our open-source project DeepFabric.

Key Responsibilities

• Research and experiment with fine-tuning strategies for language models (SFT, DPO, preference modeling, parameter-efficient methods, etc.)
• Conduct literature reviews and landscape scans to identify promising research directions relevant to SLMs
• Design controlled experiments to evaluate data quality, training strategies, and model performance
• Develop research prototypes into reproducible code assets that can integrate with our engineering pipelines
• Collaborate with engineering to translate research insights into scalable systems
• Contribute to open-source research tooling, benchmarks, and experimental frameworks
• Document findings, write research summaries, and present recommendations to the team
• Optionally collaborate on technical blog posts, internal papers, or publications

Required Qualifications

• Currently pursuing a PhD or Masters in Machine Learning, NLP, Computer Science, or a related field
• Strong Python skills and experience with ML experimentation and prototyping
• Hands-on experience with large language models, SLMs, or related research areas
• Familiarity with fine-tuning techniques (LoRA, adapters, SFT, RLHF/DPO, etc.)
• Experience with PyTorch and Hugging Face Transformers / TRL
• Ability to run structured experiments, analyze results, and communicate findings
• Strong understanding of research methodology and literature review practices
• Interest in open-source and collaborative development

Preferred Qualifications

• Experience with dataset generation, synthetic data pipelines, or data curation research
• Familiarity with evaluation frameworks, benchmark design, or data quality metrics
• Background in LLM-based data generation, prompt engineering, or preference modeling
• Experience developing reproducible research codebases
• Understanding of active learning, data selection, or filtering strategies
• Prior contributions to open-source projects
• Experience with lightweight MLOps tools for managing experiments (Weights & Biases, MLflow, etc.)

Technical Skills

• Languages: Python
• ML Frameworks: PyTorch, Hugging Face Transformers / TRL, unsloth.ai
• Experimentation Tools: W&B, MLflow, DVC, Jupyter
• Research Areas (Nice to Have): data synthesis, fine-tuning methods, SLM efficiency techniques, evaluation models

What You'll Work On

• Novel techniques for generating diverse, high-quality training datasets
• Experimental fine-tuning workflows optimized for SLMs
• Research into data quality gates, filtering strategies, and evaluation pipelines
• Explorations into alternative supervision methods and alignment strategies
• Open-source prototypes and tools that assist the broader research community

Education

• Currently either completed MSc or enrolled in a PhD program from a relevant scientific discipline with strong mathematical and computational skills, for example Computer Science, Mathematics, Physics and Engineering

This is a UK hire, all applicants must have the right to work in the UK.