Development of a skill management platform for mapping competencies, learner profiles, training programs, and professional objectives.

 Design of ontology-backed knowledge bases to represent skills, prerequisite relations, learning resources, and domain-specific competency structures.

 Integration of knowledge graph search with LLM-based fallback methods to support grounded skill search, explanation, and recommendation.

 Implementation of skill gap analysis and personalized course or training recommendations based on user context and career goals.

 Contribution to applied workflows for Forteim and BSEA around competency modeling, training alignment, and explainable AI-assisted guidance.

 Web-first prototype for organizing tasks, notes, and lightweight memory through graph-based interactions.

 Supports task-level graph views where each task can keep its own saved memory, context, and connected notes.

 Designed as a lightweight personal task and knowledge workspace with a static frontend and Node.js/Postgres backend.

 Demo: emonat.vercel.app

 Repository: emonat

 Comprehensive analysis of recommender system algorithms tailored for the vehicle domain.

 Integration of user preferences, historical interactions, and semantic understanding of vehicle features.

 Development of personalized models considering make, model, and contextual factors.

 Use of ontologies to represent complex relationships and better capture user intent.

 Research on semantic similarity estimation between vehicles beyond basic attributes.

 Development of personalized, context-aware vehicle recommendations aligned with unique user needs.

 Repository: vivo_portals

 Focus on the study of genetic and phenotypic diversity in Asian and African rice varieties.

 Challenges addressed: handling large, heterogeneous datasets stored in varied formats.

 Developed a generic integration and indexing tool to streamline data navigation, sharing, and annotation.

 Repository: Syspherice

 Developed a tool to extract concepts and terms from PDF text files.

 Used POS tagging to identify meaningful patterns in input sentences.

 Incorporated frequency analysis to detect commonly occurring terms in general context.

 Generated a triplet dataset matching the standard Subject-Verb-Object (SVO) structure.

 Repository: semanticextraction

 Combinatory Categorial Grammars (CCG) serve as a bridge between surface syntax and semantic representations.

 Introduced a novel method to generate CCG derivation trees from dependency syntax trees, focused on French.

 Developed a binarization algorithm to convert dependency tree structures into binary trees.

 Assigned lexical categories by leveraging context and dependency relations of words and POS tags.

 Generated full CCG derivations using binary structures and CCG combinatory rules.

 Evaluation performed on the 21,550-sentence French Treebank Corpus.

 Repository: FrenchCCGBank

 Used the French CCG corpus and the Groningen Treebank for English as training datasets.

 Developed a new BiLSTM+CRF neural architecture for supertagging.

 Incorporated morphosyntactic input features and feature correlation embeddings.

 Demonstrated that dependency syntax significantly improves supertagging accuracy for French.

 Results highlight the effectiveness of deep learning for inflectional languages in syntactic tasks.

 Repository: ccg_supertagging_french