Tjippies · Softwarify®
AI Research Scientist
Description
🚀 We're hiring: AI Research Scientist, Large Language Models for Software Development (PhD required)
Softwarify turns an idea into running software, built by a team of robots that a human or team of humans direct. A Virtual CEO understands the idea and hires the team: a Virtual CTO, engineers, a designer, QA and a writer. The team plans, writes, tests and reviews the work on its own together with the human counterparts with “human in the loop” mindset.
The platform is free and open source. A cloud version runs on usage credits. In both, the model bill decides how much software a team can build.
We are developing technology to bring that cost down to a fraction of a vanilla AI subscription. We are looking for the scientist who will be a part of that research.
This is a science position: you form hypotheses, run controlled experiments, publish, and see the results that hold move into the product.
Watch a full demo of the platform here: https://www.youtube.com/watch?v=9WuD5qcGo7E
🔬 What you will work on
- Model routing and cascades: which calls can a smaller model take without losing quality?
- Smaller specialised models: distillation and fine-tuning for planning, editing and review
- Context construction: what a model must see to make a correct change, in fewer tokens
- Output efficiency: edit representations and decoding that cut output tokens
- Cache-aware prompting, and planning within a fixed budget
- Evaluation: cost per accepted change over a sequence of changes, not one isolated task
🎓 What you bring
- A completed PhD in machine learning, NLP, software engineering or a related field (minimum requirement)
- Lead-author publications at venues such as NeurIPS, ICML, ICLR, ACL, EMNLP, ICSE, FSE or ASE
- Hands-on research with large language models: training, fine-tuning, inference efficiency or tool-using agents
- Experimental rigour, strong Python and PyTorch, clear written English
💡 What you get
- A working open-source platform, so a hypothesis meets a real workload
- A token-level record of every run, and a self-hosted model gateway where a new model swaps in
- Your work published and released in the open: code, models and benchmarks
📍 Location: Stockholm, Sweden (in-person)
📅 Application deadline: January 5, 2027
📩 To apply: send your CV via LinkedIn Easy Apply, include your publications.
#hiring #AIresearch #LLM #SoftwareEngineering #OpenSource