🎓 Undergraduate Researcher & Electrical Engineering Student at Universitas Tanjungpura (UNTAN), Pontianak, West Kalimantan, Indonesia (ROR:
04exz5k48)
📍 Location: Pontianak, West Kalimantan, Indonesia
💡 Bio: Undergraduate Researcher @ UNTAN | TinyML (ESP32-S3) · Semantic Web Ontologies & GraphRAG | Author of CorpusLD | ORCID: 0009-0005-2933-7779
Passionate about bridging Edge AI / TinyML hardware constraints with Deep Semantic Web Ontologies & Knowledge Graphs. Building lightweight machine learning models that run on microcontrollers (<200KB model size, <100KB RAM) alongside deterministic document intelligence and linked data architectures.
Research Focus:
- 🔋 Renewable Energy Reliability: Lightweight 1D-CNN Open-Circuit Fault Detection for Solar PV Inverters on ESP32-S3.
- 🕸️ Knowledge Graph Engineering: Dual-Layer Semantic Web Extraction, W3C Schema.org JSON-LD, Deterministic Unit Ontologies, and Bidirectional Citation Graphs.
- 🔒 Privacy-First Local RAG Systems: Section-wise Map-Reduce extraction with zero token truncation loss.
"CorpusLD: A Dual-Layer Semantic Extraction Framework and Deep Knowledge Graph Architecture for Scientific Literature with Deterministic Unit Ontology and Authority Disambiguation"
- Author: Sharrif Faqih Fajarudin (Universitas Tanjungpura)
- Persistent DOI:
10.5281/zenodo.22179715 - Interactive Paper Hub: https://sharriffajar.pages.dev/papers/corpusld/
- Living Knowledge Graph: https://sharriffajar.pages.dev/living-knowledge/
- Citation Formats: Full 1-click export for IEEE Standard, BibTeX, APA 7th Edition, and RIS (EndNote/Zotero).
🛠️ CorpusLD — Dual-Layer Academic Linked Data & Knowledge Graph Engine
| Aspect | Details |
|---|---|
| Type | Dual-Layer Academic Linked Data Extraction Engine & Deep Knowledge Graph Studio |
| Target | Unstructured Scientific Papers, Technical Reports & Patents → Schema.org JSON-LD, RDF Turtle, Neo4j Cypher |
| Innovation | 4-Tier Hybrid Parser + 5-Agent Map-Reduce + Live Authority Resolvers (ROR v2 / Wikidata / MeSH / Crossref) |
Layer 1 (Ingestion & Extraction Engine):
PDF Upload → 4-Tier Hybrid Parser (PyPDF + LlamaParse + Unstructured + Table Stitcher) →
5-Agent Map-Reduce Pipeline (Metadata, Outline, Metrics, UniversalTable, Citations) →
Universal Unit Ontology Normalization (SI, Biomedical, Energy, Compound)
Layer 2 (Semantic Graph & Linked Data Layer):
Live Authority Resolvers (ROR v2 Registry, Wikidata QID, MeSH, Crossref & OpenAlex DOI) →
Adversarial KG Conflict Detection & Graph Health Analysis →
Multi-Format Semantic Export (Schema.org JSON-LD, W3C Turtle .ttl, Neo4j Cypher .cql, BibTeX, RIS, CSL-JSON)
Key Features:
- 🤖 5-Agent Map-Reduce Pipeline: Guaranteed 0% context truncation loss across arbitrarily long papers.
- 🏛️ Live Domain Authority Linker: Dynamic ROR v2 REST lookup for global institutions + canonical Wikidata QIDs.
- 🔬 Deterministic Unit Ontology: 8 standard scientific dimensions eliminating superscript footnote collisions (
$W/m^2$ vs$[2]$ ). - 🕸️ Bidirectional Citation Lineage: Honors foundational ancestors (Berners-Lee, Bizer, LayoutLMv3, Lewis et al.) alongside topological research gap discovery.
- 🛡️ Enterprise Security Hardened: SSRF loopback defense, strict path traversal validation, and DOM sanitization.
⚡ Lightweight 1D-CNN Edge AI for Inverter Fault Diagnosis — Undergraduate Thesis
| Aspect | Details |
|---|---|
| Type | TinyML / Embedded AI Fault Diagnosis System |
| Target | Real-time Open-Circuit Fault (OCF) detection on ESP32-S3 |
| Model Specs | INT8 Quantized 1D-CNN (21.5 KB, ~14.7k params, ~8.4 ms latency) |
| Fault Classes | 6 Classes: Healthy, S1_Open, S2_Open, S3_Open, S4_Open, Multi_Fault |
Pipeline: generate_dataset.py (128-sample window) → train_model.py (1D-CNN) →
quantize_export.py (INT8 TFLite ~21.5KB) → TFLite Micro on ESP32-S3 →
IoT Telemetry via Thinger.io
┌────────────────────────────────────────────────────────────────────────────┐
│ Phase 1: Completed ✓ (Q1 - Q3 2026) │
│ • CorpusLD v3.0 Released (104 automated unit tests passed) │
│ • Official Master Preprint Published on Zenodo (DOI: 10.5281/zenodo.22179715)│
│ • Interactive Living Knowledge Graph & Lineage Engine Live │
├────────────────────────────────────────────────────────────────────────────┤
│ Phase 2: Active / Current Focus 🎯 (Q4 2026) │
│ • Self-Hosted Empirical PoC & Google Scholar Indexation Validation │
│ • Highwire Press Metadata & 100/100 Google Rich Results Verification │
│ • Scopus Q1/Q2 Manuscript Finalization (IMRAD standard) │
├────────────────────────────────────────────────────────────────────────────┤
│ Phase 3: Next Milestone 🏆 (Q1 - Q2 2027) │
│ • Full Paper Submission to Scopus Q1/Q2 Journal (IEEE Access / Elsevier) │
│ • Laboratory Hardware Testbed Dataset Acquisition (1D-CNN ESP32-S3) │
│ • Undergraduate Thesis Defense │
├────────────────────────────────────────────────────────────────────────────┤
│ Phase 4: Productionization 🌐 (Late 2027) │
│ • Institutional OJS Journal Plugins for Automated W3C JSON-LD Export │
│ • Decentralized GraphRAG & Community Knowledge Federation │
└────────────────────────────────────────────────────────────────────────────┘
🤝 Open to academic collaborations in: Edge AI, TinyML, Knowledge Graphs, Semantic Web Ontologies, and Renewable Energy Reliability.
"Democratizing AI · Lightweight Edge Intelligence · Verifiable Knowledge"