AI4X 2025 · Conference / Workshop Paper · Literature-based discovery
Literature-based Hypothesis Generation using Large Language Models
Uchenna Akujuobi et al.
Represents my current research direction: AI systems that accelerate discovery workflows.
Uchenna Akujuobi
Research Scientist, Sony AI — graph learning, LLMs, and AI for scientific discovery.
I build intelligent systems that augment human abilities: automated hypothesis generation from scientific literature, graph representation learning, and food & health intelligence.
Publications
20+ papers
Top venues
NeurIPS · WSDM · ACL
Doctorate
PhD, KAUST 2020
Current role
Sony AI · 2020–
About
Builder instincts, research discipline.
I am an AI researcher working across machine learning, graph representation learning, scientific discovery, and applied AI systems. My path has always mixed research with building: early Android apps and technical writing, KAUST doctoral work on dataset-driven scholarly search and graph learning, and now Sony AI research on automated hypothesis generation, food and health intelligence, and systems that help scientific experts move faster.
Evidence
Data and citations before conclusions. I show my work so experts can challenge it.
Systems
I often think: how components connect, where things break, and what is needed to navigate complexity.
Taste
Good research systems should be simple and solve problems, not impressive.
Research
How can intelligent systems help people?
My work spans academic search, graph learning, scientific hypothesis generation, and food-health intelligence.
AI for scientific discovery
Building systems that help researchers generate, evaluate, and navigate hypotheses by connecting evidence across literature, graphs, biological signals, and domain knowledge.
hypothesis generation · literature intelligence · knowledge graphs
Graph representation learning
Studying how entities, labels, documents, and citations interact in complex networks — including semi-supervised node classification and reinforcement-learning graph walks.
reinforcement learning walks · multi-label classification · citation networks
Food, health, and human-centered AI
Applied AI connecting sensory experience, nutrition, gastroenterology, and personalization for healthier food decisions and better expert workflows.
food AI · health signals · expert augmentation
Publications
Selected publications
Peer-reviewed work across NeurIPS, IEEE TKDE, WSDM, ICDM, ACL, ECAI, and more. Filter by theme, sort, or search by title, venue, or author.
17 publications
AI4X 2025 · Conference / Workshop Paper · Food AI and health
Gastro-Health Project: Revolutionizing Personalized Nutrition and Health Forecasting Through Integrated AI Technologies
Uchenna Akujuobi, J Yi, ME Chung, TR Besold
Artificial Intelligence Review 2024 · Journal Article · Food computing
Revisiting Named Entity Recognition in Food Computing: Enhancing Performance and Robustness
Uchenna Akujuobi, S Liu, TR Besold
Artificial Intelligence Review 2024 · Journal Article · Biomedical hypothesis generation
Link Prediction for Hypothesis Generation: An Active Curriculum Learning Infused Temporal Graph-Based Approach
Uchenna Akujuobi, P Kumari, J Choi, S Badreddine, K Maruyama et al.
ACL Findings 2024 · Conference Paper · Natural language processing
It is Simple Sometimes: A Study on Improving Aspect-Based Sentiment Analysis Performance
L Cabello, Uchenna Akujuobi
arXiv 2024 · Preprint · Medical AI
MEG: Medical Knowledge-Augmented Large Language Models for Question Answering
L Cabello, C Martin-Turrero, Uchenna Akujuobi, A Søgaard, C Bobed
ECAI 2023 · Conference Paper · Literature-based discovery
CERM: Context-Aware Literature-Based Discovery via Sentiment Analysis
JC Young, Uchenna Akujuobi
Information Sciences 2022 · Journal Article · Graph learning
Semantic Guide for Semi-Supervised Few-Shot Multi-Label Node Classification
L Xiao, P Xu, L Jing, Uchenna Akujuobi, X Zhang
IEEE TKDE 2022 · Journal Article · Biomedical hypothesis generation
T-PAIR: Temporal Node-Pair Embedding for Automatic Biomedical Hypothesis Generation
Uchenna Akujuobi, M Spranger, SK Palaniappan, X Zhang
One of my most-cited papers (32 citations) — a core contribution to temporal hypothesis generation.
KAUST 2020 · PhD Thesis · Scholarly search
Learning from Scholarly Attributed Graphs: A Graph Representation Learning Approach
Uchenna Akujuobi
NeurIPS 2020 · Conference Paper · Biomedical hypothesis generation
Temporal Positive-Unlabeled Learning for Biomedical Hypothesis Generation via Risk Estimation
Uchenna Akujuobi, J Chen, M Elhoseiny, M Spranger, X Zhang
Published at NeurIPS 2020 — the highest-profile venue in my publication record.
WSDM 2020 · Conference Paper · Graph learning
Recurrent Attention Walk for Semi-supervised Classification
Uchenna Akujuobi, Qiannan Zhang, Han Yufei, Xiangliang Zhang
A key graph learning contribution from the KAUST research period, published at WSDM 2020.
ICDM 2019 · Conference Paper · Graph learning
Collaborative Graph Walk for Semi-supervised Multi-Label Node Classification
Uchenna Akujuobi, Han Yufei, Qiannan Zhang, Xiangliang Zhang
ICDM 2019 · Conference Paper · Dataset recommendation
Dataset Recommendation via Variational Graph Autoencoder
B Altaf, Uchenna Akujuobi, L Yu, X Zhang
Most-cited of my papers (44 citations) — a strong result on dataset-driven scholarly AI.
IEEE BigData 2018 · Conference Paper · Dataset mining
Mining Top-k Popular Datasets via a Deep Generative Model
Uchenna Akujuobi, Ke Sun, Xiangliang Zhang
ACM SIGKDD Explorations 2017 · Journal Article · Scholarly search
Delve: A Dataset-Driven Scholarly Search and Analysis System
Uchenna Akujuobi, Xiangliang Zhang
Bridges academic search, dataset retrieval, and visual analysis — built during my KAUST doctoral period.
ECML PKDD 2017 · Conference / Demo Paper · Scholarly search
Delve: A Data Set Retrieval and Document Analysis System
Uchenna Akujuobi, Xiangliang Zhang
Projects
Selected work
Each project is framed by problem, system, and outcome — from current Sony AI research to early products.
Scientific AI · 2025
Automated Hypothesis Generation
A current research direction around multi-agent systems that read, connect evidence, and produce scientifically useful hypotheses.
Positions AI as a research collaborator: less autocomplete, more structured scientific reasoning.
multi-agent systems · LLMs · scientific discovery · in progress
Scholarly search · 2017–2019
Delve
An academic search engine for dataset retrieval and document analysis, visualizing relationships among papers, citations, and datasets.
Turned dataset discovery into an interface problem: helping researchers locate benchmark data and understand how it is used.
scholarly search · citation networks · dataset retrieval · sucessful
Food and health AI · 2024
Gastro-Health
A food recommendation research project incorporating gastrointestinal awareness instead of relying only on taste or preference signals.
A more human model of recommendation: food decisions connected to health context.
recommendation · food AI · health · in progress
Game AI · 2017
Mancala3D
An African Mancala-inspired game project connected to earlier work comparing search algorithms for game-playing agents.
A personal bridge between culture, software craft, and search algorithm experimentation.
game AI · Android · search algorithms
Community tooling · 2017
KAUST Orientation App
A mobile app with information and tools for new students joining the KAUST community.
A practical example of product-minded engineering: reduce friction for real people in a specific context.
Android · student experience · open source · handed over
Commerce mobile · 2016
TenSold Android App
An Android app developed for the TenSold online shop — an early mobile development project outside academia.
Early evidence that I could ship real, user-facing software alongside academic work.
Android · commerce · mobile development · handed over
Startup · 2016
GINIPROX
A project and idea collaboration platform, documented candidly as a failed project and a real product lesson.
A useful product lesson: technical ambition needs market research, user understanding, and clear positioning.
website · project management · startup · collaboration · failed project
Experience
Experience
2020 – present
Tokyo, Japan; Barcelona, Spain
Senior Research Scientist
Sony AI
I research AI systems for scientific discovery, food intelligence, recommendation, and knowledge-rich expert workflows at the intersection of graph learning and generative AI.
- Research directions include automated hypothesis generation and multi-agent systems for scientific literature.
- Contributing to food and health AI, including the Gastro-Health recommendation project and a nutrition knowledge graph covering 1,000 foods.
- Working at the boundary of graph learning, large language models, and applied product research.
2016 – 2020
Thuwal, Saudi Arabia
PhD Researcher
King Abdullah University of Science and Technology (KAUST)
Doctoral research in machine learning and data science, spanning dataset retrieval, citation network analysis, and graph representation learning.
- Built Delve, a dataset-driven scholarly search and analysis system published in ACM SIGKDD Explorations and demonstrated at ECML PKDD.
- Published graph walk methods for semi-supervised classification at WSDM 2020 and ICDM 2019.
- Dataset mining work published at IEEE BigData 2018.
2021 – 2022
Tokyo, Japan
Research Intern
Sony Computer Science Laboratories
Research internship bridging KAUST doctoral work and the Sony AI position, applying academic machine learning methods to product-oriented research questions.
- Connected graph learning and scientific discovery research to applied product contexts.
- Built foundations for the food and health AI research agenda that continued at Sony AI.
Writing
Notes and tutorials
Sony AI / research
AI for scientific hypothesis generation
A current research thread on why I care about systems that help experts reason across literature and evidence.
Read note →Legacy UCAKU · 2017
Setting up a web bot using Tor
A practical guide to configuring a privacy-preserving web bot with Tor for legitimate research, testing, and data collection, including proxy setup, request routing, rate limiting, and responsible usage practices.
Read note →Legacy UCAKU · 2017
Configuring Elasticsearch 5.3 on CentOS 7
Step-by-step setup of Elasticsearch, Kibana, X-Pack, and Logstash on CentOS 7
Read note →Legacy UCAKU · 2017
Configuring Elasticsearch on Ubuntu 14.04
The Ubuntu version of the Elasticsearch stack setup covering Elasticsearch, Kibana, X-Pack, and Logstash on Ubuntu 14.04.
Read note →Graph learning · 2020
Why graph walks work for semi-supervised learning
A short note on the intuition behind random walk methods for node classification and why the graph structure matters more than the labels.
Read note →Delve · KAUST · 2018–2019
What I learned building Delve
Reflections on building a dataset-driven scholarly search engine: the interface problems that shaped it, and what it taught me about research infrastructure.
Read note →Contact
Get in touch
Reach out about research collaboration, scientific AI systems, food intelligence, graph learning, or thoughtful product work around expert workflows.