AI/ML Engineer and Engineering Manager specialising in production LLM systems, Transformer fine-tuning, and end-to-end ML deployment.
I like building things that actually work in production — not just in notebooks.

• LLM & Agentic Systems — production pipelines, context engineering, and NLP-powered features built with LangChain/LangGraph and FastAPI
• Transformer Fine-tuning — classification, seq2seq translation, and encoder-decoder architectures across the full pipeline from data prep to serving
• Computer Vision — image classification, object detection, and OCR for real-world applications
• ML Infrastructure — model deployment, monitoring, and evaluation with MLflow, LangSmith, Docker, and Kubernetes

🧠 Neurosymbolic LM — Exploring improved reasoning in language models via a T5-Gemma backbone augmented with lightweight intermediate Transformer modules. These extract entity and relationship structure from inputs and broadcast to global attention, giving the decoder structured world knowledge during generation.

🛡️ Image Immunization — Adversarial image protection against misuse in diffusion pipelines. A NestedUNet trained to generalise perturbations that disrupt VAE encoding, guided by the DiffVax approach — giving creators more control over their IP.

Python · PyTorch · Hugging Face Transformers · LangChain · LangGraph · FastAPI · MLflow · LangSmith · vLLM · Elasticsearch · Neo4j · Docker · Kubernetes

💡 Chronic ideas person · 📐 BSc Software Engineering, University of Portsmouth · 📬 sam.bartlett858@gmail.com