Featured
Seismic Inpainting with Deep Priors
Reconstructing missing seismic traces using U-Net + implicit neural representations for subsurface imaging.
PyTorch U-Net INR NumPy
Read case study
Available for consulting — 2 slots open
Ridhwan Amin
AI/ML Researcher & Engineer
I build seismic inpainting models, anomaly detection systems, and hybrid RAG pipelines at UTP. I also consult with teams that need production-ready ML infrastructure.
01 / Projects
Featured work
Hybrid RAG Pipeline for Enterprise Docs
Combining dense retrieval and keyword search with an LLM reranker to answer questions over large document corpora.
LangChain FAISS OpenAI FastAPI
Read case study
Oilfield Anomaly Detection System
Real-time multivariate anomaly detection on SCADA sensor streams using Isolation Forest + LSTM autoencoders.
scikit-learn LSTM Kafka Grafana
Read case study
LLM Evaluation Benchmark Suite
Automated benchmark suite for evaluating instruction-tuned LLMs on domain-specific scientific reasoning tasks.
Python HuggingFace Pandas Weights & Biases
Read case study
02 / About
MEXT Scholar 2022
IEEE Member
UTP Researcher
M.Sc. EEE
PETRONAS Digital Alumni
About me
I'm Ridhwan Amin, an AI/ML researcher at Universiti Teknologi PETRONAS (UTP) in Malaysia. My research sits at the intersection of deep learning and geophysical signal processing — specifically, using implicit neural representations and generative priors to reconstruct missing seismic data.
Beyond research, I consult with teams that need production-ready ML infrastructure — RAG pipelines that actually retrieve the right things, anomaly detection systems that don't cry wolf, and MLOps setups that let your team iterate without fear.
I received a MEXT scholarship from the Japanese government and am a member of IEEE. I also write about ML, research craft, and working in energy-domain AI on my Substack .
03 / What I Build
Capabilities
ML Research & Modelling
Designing and training deep learning models for scientific and industrial applications — from seismic signal processing to NLP. Publication-grade rigour with production feasibility in mind.
PyTorch JAX scikit-learn CUDA
RAG & LLM Systems
Building retrieval-augmented generation pipelines that go beyond naive chunking. Dense + sparse hybrid retrieval, reranking, and evaluation harnesses for enterprise document Q&A.
LangChain LlamaIndex FAISS pgvector
MLOps & Infrastructure
Taking models from Jupyter notebooks to reliable, monitored production systems. Experiment tracking, CI/CD for ML, model registries, and inference optimisation.
MLflow DVC Docker FastAPI
Data Engineering
Designing and implementing data pipelines that feed ML systems with clean, well-labelled data — including streaming sensor data, geospatial datasets, and scientific corpora.
Spark dbt Airflow Kafka
04 / Journey
Timeline
- 2023–now2023–now
AI/ML Researcher
Universiti Teknologi PETRONAS (UTP)Leading research on seismic inpainting with deep generative priors. Collaborating with PETRONAS on subsurface imaging pipelines.
- 2022–20232022–2023
MEXT Research Scholar
Japanese Government Scholarship
Government-funded research fellowship supporting AI/ML work in energy-domain applications.
- 2021–20232021–2023
M.Sc. Electrical & Electronic Engineering
Universiti Teknologi PETRONAS (UTP)
Thesis: Seismic trace reconstruction using implicit neural representations and energy-based priors.
- 2020–20212020–2021
ML Engineer (Intern)
PETRONAS Digital
Built anomaly detection prototypes for oilfield SCADA data. Introduced MLflow experiment tracking to the team.
- 2017–20212017–2021
B.Eng. Electrical & Electronic Engineering
Universiti Teknologi PETRONAS (UTP)
First-class honours. Final year project on signal reconstruction with compressed sensing.
05 / Research
Publications
Accepted — pending publication
Deep Prior-Based Seismic Trace Inpainting Using Implicit Neural Representations
IEEE Transactions on Geoscience and Remote Sensing · 2024
Ridhwan Amin, Mohd. Hafiz Hashim, Ahmad Fadzil M. Hani
We propose a method combining implicit neural representations (INRs) with energy-based deep priors to reconstruct missing seismic traces without requiring labelled training data. Our approach achieves 3.2 dB PSNR improvement over conventional interpolation baselines on the SEAM Phase I dataset.
DOI: 10.1109/TGRS.2024.XXXXXXXHybrid Retrieval-Augmented Generation for Scientific Literature: A Domain Adaptation Study
IEEE Access · 2023
Ridhwan Amin, Nurfadzilah Ahmad
This paper evaluates hybrid retrieval strategies — combining dense embeddings with BM25 keyword retrieval — for question-answering over petroleum engineering literature, demonstrating a 12% improvement in answer faithfulness over dense-only baselines.
DOI: 10.1109/ACCESS.2023.XXXXXXX06 / FAQ
Frequently asked
I focus on ML/AI projects in the scientific and industrial domain — seismic signal processing, NLP/RAG pipelines, anomaly detection, and MLOps. I'm most useful when there's a real research problem to solve, not just a model to fine-tune. See the Services page for pricing tiers.
I'm currently based at UTP as a researcher, so I take on consulting and part-time advisory engagements rather than full-time roles. If you have something compelling, reach out and we can discuss.
I sign NDAs as standard for consulting work. For sensitive datasets, I can work in a sandboxed environment you provide, or on anonymised/synthetic data. We discuss data handling in the initial scoping call.
It starts with a free 30-minute scoping call. I then send a written proposal with scope, deliverables, timeline, and fixed price. Work proceeds in 2-week sprints with a check-in call each cycle. Final deliverables include code (well-documented), a technical report, and a handover session.
Yes — I'm open to research collaborations, co-authoring, and advisory roles on ML-heavy research projects. Drop me an email with the research question and your institution.
05 / Contact
Let's work together
Whether you need ML infrastructure, a research collaboration, or just want to talk about seismic imaging — I'm reachable.
-
Email
ridhwan@ridhwan.dev -
LinkedIn
linkedin.com/in/ridhwanamin -
GitHub
github.com/RidhwanAmin