Applied Machine Learning Engineer
Ahmad Royyan Fatah
Physics undergraduate at ITB building production-oriented machine learning systems, intelligent data products, and LLM-powered applications across predictive modeling, computer vision, and document analysis.
- At Artajasa, built 4 containerized Go and Python microservices, benchmarked 8 facial-recognition models on 2,000 image pairs, and delivered a live system using gRPC, Protocol Buffers, and pgvector.
- At FTTM ITB, benchmarked 5 models across 4 scenarios, improved hydrocarbon recall by 45%, reached 0.71 F1-score and 0.87 AUC-ROC, and recorded RAPIDS cuML speedups up to 32.7x.
- Built Kuliah Kit with multimodal ingestion for 5 file types, structured LLM outputs, concept-tagged quiz analytics, KaTeX rendering, and feedback instrumentation for iterative improvement.
Interested in work that connects applied machine learning, data systems, and product delivery.
Featured Projects
Selected work across applied machine learning, intelligent product development, and full-stack system delivery.
About Me
I study Physics at ITB with a concentration in Science Data and Artificial Intelligence, and I work across applied machine learning, full-stack product development, and technical instruction. Recent work includes hydrocarbon modeling research, computer-vision systems for banking, and education platforms built with modern web and data stacks.
- Feature engineering, experimentation, model evaluation, and class-imbalance handling
- Real-time inference, vector search, and structured LLM output generation
- Python, TensorFlow, PyTorch, RAPIDS cuML, PostgreSQL, and Docker
- Next.js, TypeScript, Go, Flutter, PostgreSQL, and tRPC
- Ownership across architecture, APIs, inference services, and product integration
- Experience with payments, subscriptions, content systems, and education platforms
- Benchmarking and reproducible experimentation in scientific and product settings
- Executive-facing delivery for high-stakes demos and institutional systems
- Teaching and curriculum design for Python and analytical workflows
Professional Experience
Recent roles center on applied machine learning, production-style AI systems, fintech strategy, and domain-specific instruction.
Developed business cases and financial simulations for Web3 remittance, treasury, and MSME financing use cases in Indonesian institutional contexts.
- Modeled a stablecoin settlement scenario showing 91% cost reduction for cross-border remittance
- Evaluated ISO 20022 middleware concepts, RWA revenue models, and decentralized e-factoring opportunities
Applied machine learning to low-resistivity well-log interpretation and cross-well hydrocarbon prediction, combining domain features, class-imbalance handling, and GPU benchmarking.
- Benchmarked 5 models across 4 scenarios, engineered 7 petrophysical features, and improved hydrocarbon recall by 45%
- Reached 0.71 F1-score with 0.87 AUC-ROC and recorded RAPIDS cuML speedups up to 32.7x
Independently built Bersama Bill ID, a production-style biller-identification proof of concept with facial authentication, payment orchestration, and a mobile client for live executive demonstration.
- Architected 4 containerized Go and Python microservices with gRPC, Protocol Buffers, PostgreSQL, and pgvector
- Built the Flutter client and benchmarked 8 recognition models before the live demo to banking executives
Designed and delivered an intensive tutorial for petroleum engineering students on Python-based well-log analysis workflows.
- Covered LAS parsing, Pandas and NumPy workflows, signal smoothing, and multi-track visualization
- Bridged geoscience interpretation with modern data analysis practices in hands-on sessions
Leadership Experience
Beyond internships and research, much of the execution work has been in student organizations where the scope was effectively that of real product and platform ownership.
Leading PHIWIKI's shift from print-only educational products toward digital subscriptions and platform-based learning.
- Built the online course platform with structured content, quiz attempts, progress tracking, and S3-backed media delivery
- Integrated Mayar payments, referral coupons, and role-based access control for premium access
Architected and launched the official Physics Student Association website as a shared digital platform for content, identity, and branded utilities.
- Built the News and Blog system, branded shortener with QR export, and admin user management
- Aligned the RBAC and infrastructure model with the PHIWIKI platform ecosystem
Led both the website build and the digitization of G-Corp's flagship exam-prep books for incoming students.
- Built the official site, branded shortener, and multimedia content workflows on a type-safe full-stack architecture
- Migrated BSG production from Word to LaTeX and embedded QR-linked explanations into the printed books
Built the event operations platform from zero to production in a two-week sprint to support 100+ participants.
- Implemented task distribution, attendance and leave workflows, daily condition tracking, and admin-side scoring
- Integrated Google Sheets synchronization so non-technical organizers could operate from familiar spreadsheets
Education & Training
Built a quantitative foundation through physics, computation, experimentation, machine learning, data visualization, instrumentation, and product-oriented coursework.
Completed a 463-hour curriculum spanning Python automation, data analysis, classical machine learning, TensorFlow, deployment, and generative AI, graduating with a 90.2% final mark.
Taught Python fundamentals, debugging, algorithmic problem solving, data analysis, and domain-specific analytical workflows to first-year students and petroleum engineering participants.
Technical Skills
Other Projects & Experiences
While my recent work focuses on bigger things, these projects and experiences represent my early-stage work that preceded my current roles. From teaching others to delivering early-stage software, these building block experiences remain central to my ability today.
Roles
Guided first-year computation students in algorithmic problem-solving, Python programming, and computational thinking.
Maintained full-stack web solutions to strengthen the organization's digital presence and member-facing delivery.
Built responsive Next.js and React interfaces in a cross-functional delivery environment for client-facing web work.
Developed user-facing interfaces for executive-board operations and student engagement initiatives.
Supported web-based event operations and participant-management systems for a nationwide physics competition.
Handled data collection and information workflows for the student association election process.
Produced worked examples and physics learning materials that became part of PHIWIKI's printed products.
Projects
AI-powered study workflow for lecture materials with multimodal ingestion, structured summaries, learning goals, quizzes, KaTeX rendering, and feedback instrumentation.
Habit-tracking application built around loss-aversion and streak mechanics to improve consistency and retention.
Multi-tenant file manager for S3-compatible storage with a modern browser-based interface.