Linlin Jia, Ph.D.
Machine Learning Research ScientistAdvanced Postdoc @ University of Bern
Graph Machine Learning · Spatio-Temporal Learning · Graph AI for Science & Industry · LLM Systems & Agents
Open to Work
Actively applying — available in short time
News & Updates
About Me
I am an Advanced Postdoctoral Researcher at the University of Bern, Switzerland, working at the Pattern Recognition Group. My expertise spans graph representation learning, spatio-temporal deep learning, and LLM-based AI systems.
I received my Ph.D. in Computer Science in 2021 from the LITIS Lab, INSA Rouen, Normandy University, France, under the supervision of Prof. Paul Honeine and Prof. Benoit Gaüzère, focusing on graph machine learning and pattern recognition in chemoinformatics. Since then, I've contributed to projects in graph ML for computational chemistry (polymer optimization, redox prediction), document analysis (historical papyri, engineering drawings), and environmental science (river-temperature forecasting). I've supervised 20+ students and contributed to multiple scientific grants. Prior to that, I earned my M.Sc. in Software Engineering (2017) and B.S. in Information Engineering (2014), both from Xi'an Jiaotong University, China.
Beyond academia, I collaborate with N-Banker, a FinTech startup, where I lead AI strategy and LLM-agent development. I have also built and contributed to multiple open-source libraries, ML toolkits, and LLM-agent systems.
I am actively seeking ML Research Scientist / Engineer opportunities in both academia and industry, with a focus on graph-based learning, spatio-temporal learning, scientific computing, and AI-driven discovery and industrial applications.
Research Outline
Graph Machine Learning
Bridging graph kernels, edit distances, GNNs, transformers. Pre-image and generation problems.
Spatio-Temporal Learning
Time series with entity embeddings, spatial graphs for environmental forecasting.
Graph AI for science and industry
Computational chemistry, smart engineering, hydrology, digital humanities, finance, healthcare.
LLM Systems & Agents
RAG pipelines, fine-tuning (LoRA), agent orchestration for production AI.
Professional Experience
Advanced Postdoctoral Researcher
SNSF-funded research on spatio-temporal graph convolutional networks and LLMs for Swiss river temperature forecasting. Developing transformer architectures with entity embeddings and graph structures. Paper accepted by ICPR 2026.
Scientific Collaborator
Applied topology-aware graph and vision deep learning for symbol detection, connection prediction, and graph construction in engineering diagrams and handwritten document analysis. Contributed to SNSF-funded Virtual Bodmer project (accepted) for 3D digital reconstruction of ancient papyrus.
Research Fellow
SNSF-funded research on novel graph matching algorithms. Published at ACPR 2023 on bridging distinct spaces in graph-based machine learning.
Postdoctoral Researcher
Applied graph ML to polymer optimization (OCTOPUSSY project) and redox potential prediction. Published in Journal of Computational Chemistry (2024).
Ph.D. in Computer Science
Thesis: "Bridging Graph and Kernel Spaces: A Pre-image Perspective" — Graph kernels, graph edit distances, and pre-image problems.
Projects
LIULIAN — Liquid Intelligence and Unified Logic for Interactive Adaptive Networks
- Spatio-temporal data platform: collection, processing, analysis, and visualization for scientific and engineering domains.
- Multimodal scientific graph benchmarks.
- LLM-driven data interfaces and BI agent systems.
- Consolidates PhD/postdoc research in graph ML (kernels, GED, GNNs, pre-image).
- Preparing Innosuisse Innovation Project funding application.
Spatio-Temporal Graph Convolutional Networks for River Temperature Forecasting
- SNSF-funded project on graph-based Swiss river water temperature forecasting.
- Developed spatio-temporal transformer and LLM architectures with entity embeddings + graph structures.
- Paper accepted at ICPR 2026.
- 🤝 Partners: Pattern Recognition Group (UniBE) × Berne UAS (BFH).
N-Banker: 1st Global Neobank Research Center
- Leading AI strategy and LLM agent development for the Neobank Research Center platform.
- Automated complex information retrieval, analysis, and strategic consultation for business partners.
- Built on machine learning + LLM stack (RAG, agents, FinTech pipelines).
- 🤝 Partners: Digital Financial Services Research Center × Hong Kong Polytechnic University.
Combining Image and Graph-based Neural Networks for Handwriting Recognition
- SNSF-funded project on handwritten historical document analysis.
- Combines image-based sematic with graph-based topology for deep learning.
- Paper under revision.
- 🤝 Partners: iCoSys Institute (HEIA, HES-SO Fribourg) × TU Dortmund.
PLANALYSER — Automated HVAC-Concept Audit and Optimisation using AI
- Innosuisse-funded AI-driven HVAC-concept audit and optimization on engineering drawings.
- Contributions: data preprocessing, model development (symbol / edge / topology extraction), pipeline design.
- 🤝 Partners: iCoSys Institute (HEIA, HES-SO Fribourg) × WATTELSE AG.
Novel State-of-the-Art Graph Matching Algorithms
- SNSF-funded research on novel graph matching algorithms.
- Bridges graph kernels, graph edit distance, and GNN embedding spaces.
- Published at ACPR 2023: Bridging Distinct Spaces in Graph-based Machine Learning.
- Host: Pattern Recognition Group, University of Bern.
OCTOPUSSY — Optimization of Polymers Using Sustainable SYnthesis
- Graph-based machine learning and chemical descriptor design for polymer optimization and redox-potential prediction.
- Published in Journal of Computational Chemistry (2024).
- RedoxPrediction package open-sources the implementation.
- 🤝 Partners: COBRA × LITIS (Univ. Rouen) × CSCT (Univ. Bath).
APi — Apprivoiser la Pré-image
- ANR-funded thesis research on the pre-image problem in ML for structured data.
- Focus: pattern recognition in discrete structured spaces.
- Produced 3 journal articles, 2 workshop papers, and the graphkit-learn library.
- Host: LITIS lab (Université Rouen Normandie).
Service-oriented Programmable Control and Scheduling for Software Defined Network
- M.Sc. research on service-oriented programmable control for SDN.
- Applied extreme learning machine (ELM) to predict network mobility in Named Data Networks.
- Resulted in one China patent (CN106376041B).
- Host: Xi'an Jiaotong University.
Local Confidential Translator
- Fully offline, privacy-first document translation powered by local LLMs.
- MIT-licensed open source, deployable via a single Docker command.
- Built end-to-end via AI-assisted (vibe) coding; MVP1 in testing.
Selected Publications
130+ citations · Google Scholar
Bridging Graph and Kernel Spaces: A Pre-image Perspective
Graph kernels, graph edit distances, and pre-image algorithms and applications. Supervised by Prof. Paul Honeine and Prof. Benoit Gaüzère.
Academic Services
Supervision
- 20+ students supervised
- Topics: Computer vision, Graph-based learning, Smart engineering, Deep learning, LLMs, Agent systems
Associations
- Member of the Swiss Association for Pattern Recognition (SAPR) (2024-)
- Member of the Marie Curie Alumni Association China Chapter (2024)
- Associated member of the LITIS lab (2022)
Reviewing
- International Conference on Pattern Recognition 2024
Skills & Technologies
Contact
Let's Connect
Location
University of Bern · Pattern Recognition Group
Schützenmattstrasse 14, 3012 Bern, Switzerland
Phone
+41 78 224 1419 (CH)
Visitors
— total visits — countries past 3 days via Microsoft Clarity