I specialize in machine learning for real-world sensor data. A single
distributed acoustic sensing interrogator can produce
more than a gigabyte
of raw data per second; most of my research concerns turning
measurements at that scale into decisions, from leak detection in
pipelines to computer vision for subsea inspection. Recent directions
include physics-informed methods for energy systems and large language
model tooling, with research partnerships across Europe and
Southeast Asia.
I also keep a public lab notebook of my hands-on LLM research:
notes.hasanarief.dev
Latest from the notebook
Research interests
Machine Learning · Deep Learning · Computer Vision ·
Fiber-Optic & Seismic Sensing · Physics-Informed Neural Networks ·
Knowledge Distillation · LLMs & Tool-Calling Agents ·
3D Point Clouds · Healthcare Technology
Current work
- AQUAROM: disease management in aquaculture
- Seven-institution consortium across Norway, the Philippines, and Indonesia:
can smart algorithms with minimal sensors match expensive monitoring for
early fish-disease detection? Includes three PhD positions building
research capacity in aquaculture-dependent regions.
- LLM tooling for data exploration & digital twins
- Chat-driven interfaces where models call tools from a constrained catalog
(Model Context Protocol) and return typed, auditable outputs: locally
hosted Mistral for text, Qwen Vision for multimodal. Related work on
detecting and reducing LLM hallucinations.
- Distributed sensing for infrastructure & the subsurface
- ML on distributed acoustic and temperature sensing (DAS/DTS) for pipeline
leak detection and multiphase flow estimation; passive seismic fiber
sensing for groundwater estimation (S-TRANET). Built on PyTorch and
ObsPy, running on HPC (high-performance computing) resources.
- Computer vision for offshore inspection
- Real-time detection and classification in subsea imagery, using
RT-DETR (Real-Time Detection Transformer) for detection and SAM-3
(Segment Anything Model) for weak-supervised annotation, feeding a
live streaming inspection pipeline with industry partners.
- Physics-informed ML & HPC (Norwegian AI Cloud)
- Physics-informed neural networks for energy systems (green hydrogen
electrolyzers) and climate teleconnection analysis, with hybrid
optimization combining a deterministic crowding genetic algorithm
with CMA-ES (covariance matrix adaptation evolution strategy).
Datasets published openly.
- NOR-DMT: digital music therapy
- A platform for virtual music therapy on GDPR-compliant (General Data
Protection Regulation) infrastructure, combining real-time video with
machine learning analysis of patient engagement, producing structured
clinical reports for the therapists.
Selected publications
- Nonlinear interpolated variational autoencoder for generalized fluid content estimation. Geoenergy Science and Engineering, 2025
- From mesh to neural nets: multi-method evaluation of PINN and Galerkin FEM for nonlinear convection–reaction–diffusion. Int. J. Applied and Computational Mathematics, 2025
- Better modeling out-of-distribution regression on distributed acoustic sensor data using anchored hidden state mixup. IEEE Trans. Industrial Informatics, 2022
- Towards building a distributed virtual flow meter via compressed continual learning. Sensors, 2022
- A survey on distributed fibre optic sensor data modelling and ML for multiphase fluid flow estimation. Sensors, 2021
- SAnE: Smart annotation and evaluation tools for point cloud data. IEEE Access, 2020
- Addressing overfitting on point cloud classification using Atrous XCRF. ISPRS J. Photogrammetry & Remote Sensing, 2019
- Land cover segmentation of airborne LiDAR data using stochastic atrous network. Remote Sensing, 2018
Full list on Google Scholar →
Education
- Ph.D. in Applied Informatics
- Norwegian University of Life Sciences, 2020
- M.Sc. in Informatics
- Bandung Institute of Technology, 2016
- B.Sc. in Computer Science
- Sepuluh Nopember Institute of Technology, 2011
Links