Location: Remote
Duration: 2–4 months (project-based)
Type: Contract / Research Collaboration (Paid)
About the Project
We are looking for a Master’s or PhD student to work on fine-tuning large language models (LLMs) for domain-specific tasks. The goal is to take an existing pretrained model (e.g., Meta AI’s LLaMA-class models or similar) and specialize it for a narrow, high-value use case using efficient fine-tuning techniques.
This is a hands-on applied project designed for someone who wants real-world experience deploying and optimising LLM systems.
Help drive the next wave of applied AI by demonstrating how fine-tuned LLMs can unlock advanced, real-world use cases beyond general-purpose foundation models. Organizations that require domain-specific accuracy, self-hosted deployments, customisable workflows, or performance beyond out-of-the-box capabilities increasingly rely on fine-tuned models to meet those needs.
Through this project, you will contribute to building specialised AI systems that deliver improved accuracy, efficiency, and control compared to out-of-the-box models. You will also help bridge the gap between academic knowledge and real-world application by applying fine-tuning techniques to solve concrete business problems.
What You’ll Work On
Fine-tuning pre-trained LLMs on small to medium datasets (500–20k examples)
Implementing parameter-efficient fine-tuning (e.g., LoRA-style methods)
Optimising training for cost and performance
Running experiments on GPU cloud infrastructure
Evaluating model performance and tradeoffs (specialisation vs generalisation)
Deploying fine-tuned models for inference
Experience
Strong Python skills
Experience with deep learning frameworks: PyTorch (preferred) or TensorFlow
Experience with Hugging Face Transformers or similar ecosystems
Hands-on experience training or fine-tuning transformer models on GPUs (local or cloud-based)
Previous experience using cloud platforms for model training or deployment (e.g., AWS, GCP, Azure, RunPod or similar GPU providers)
Experience working with or fine-tuning open-weight LLM families (Gemma-3, Qwen-3.5, Llama 4, GPT-OSS, Mistral...)
Hands-on experience with LoRA
Understanding of:
Fine-tuning vs pretraining
Overfitting and generalization
Model evaluation
Strong business awareness: ability to understand the context of the fine-tuning task and translate domain requirements into clear modeling objectives
What you bring
MSc or PhD student in Computer Science, Machine Learning, AI, or related field
Alternatively, 6 months of hands-on experience training and fine-tuning deep learning models
Has worked on LLMs in research or industry
Has fine-tuned at least one transformer model
Comfortable working independently
Interested in applied AI and real-world constraints (cost, latency, memory)
What You’ll Gain
Real-world experience fine-tuning large models (30B–100B parameter class)
Exposure to production constraints and deployment
Opportunity to co-author technical writeups if applicable
Strong applied portfolio project
What We Offer
100% Remote Work: Work from anywhere with flexibility and autonomy
Dynamic, High-Impact Projects: Work on cutting-edge ML and GenAI solutions across diverse industries
International Clients: Collaborate with global organizations and solve real-world challenges at scale
Urban Sports Club Membership: Supporting your physical and mental wellbeing
Monthly Bolt Credits: For rides
Company Events & Offsites: Regular team gatherings to connect, collaborate, and celebrate