I build interactive systems for decision support, communication and learning. My background and training are in speech, language(s) and cognition. I now combine this with software engineering, applied AI and Human-Computer Interaction: conducting empirical research, developing and testing prototypes and building practical tools.

• Interactive systems for communication, learning and decision support
• Applied AI workflows and model behaviour
• Multilingual and speech/language AI
• Full-stack applications for complex data and human judgement

Human-in-the-loop AI workflow for turning messy communication threads into structured case analysis and response drafts.

Demonstrates: LLM workflows, evaluator-optimizer loops, rubric-based evaluation, case-state extraction, draft revision and human-guided AI assistance.

Tech: Python, LangGraph, Claude API, Pydantic, Streamlit
Repo: https://github.com/sjtobin/lexia

Decision-support platform for solar power plant planning using satellite data and openEO APIs.

Demonstrates: full-stack product engineering, API orchestration, geospatial data, interactive maps, applied AI interfaces and decision-support UX.

Tech: Next.js, TypeScript, FastAPI, Python, Docker, PostgreSQL, Redis, openEO
Repo: https://github.com/sjtobin/SolarRanger

Reproducible R analysis of Afrikaans nasal airflow and eye-tracking data from speech research.

Demonstrates: empirical research, statistical modelling, speech/language data, reproducible workflows and human-subject analysis.

Tech: R, GAMMs, fPCA, acoustic analysis, eye-tracking
Repo: https://github.com/sjtobin/afrikaans-nasalization-analysis

Applied ML work on MR brain scan segmentation using 2D and 3D U-Nets.

Demonstrates: PyTorch, medical imaging, segmentation models and ML evaluation.

Tech: Python, PyTorch, medical imaging
Repo: https://github.com/sjtobin/brain_tumor_segmentation

Early-stage concept for simulation and assessment of high-stakes professional communication.

The idea is to help technical, academic and international professionals practise explaining, defending and reframing complex work under pressure.

Core loop: attempt → structured diagnosis → targeted retry → comparison/improvement.

PhD in Language & Cognition. Research experience (experimental and theoretical) across speech perception and production, phonetic adaptation, HCI, VR, multilingual communication and human-subject experiments. Software engineering training through 42 Berlin and current computational science work at Universität of Potsdam.

• LinkedIn: https://linkedin.com/in/stephen-tobin-4b18404
• ORCID: https://orcid.org/0000-0002-8229-289X
• Google Scholar: https://scholar.google.com/citations?user=2NhUbfEAAAAJ
• GitUP: https://gitup.uni-potsdam.de/tobin