LLM & GenAI Systems
Practical AI workflows using RAG, prompt orchestration, structured outputs, and model evaluation patterns that prioritize correctness and usefulness.
AI Engineer & Data Scientist · Indonesia
Muhammad
Nuril Huda
Building practical AI systems from data, models, and real-world workflows.
I turn messy data and manual workflows into AI products that people actually use. My evaluation background means I test model behavior before I trust it.
- M.Cs. Computer Science, UGM
- Founder, Aureum
- LLM Evaluation · Alignerr & Outlier
- Python · FastAPI · PySpark · Docker · AWS · GCP
About
Ambiguous problems in. Working systems out.
I work where AI engineering, data science, and product execution meet. Give me an operational problem that arrives half-formed, and I'll turn it into a structured system: define the workflow, reason about the data, evaluate how the model behaves, then ship something people can actually use.
I've worked on LLM evaluation, AI products, forecasting pipelines, backend RAG applications, and medical computer vision research. The common thread: I care whether the system actually helps someone decide or act, not whether the demo looks good.
What I build
Four kinds of systems.
AI Product Engineering
Backend-driven AI applications with Python, FastAPI, Docker, relational workflows, caching, and cloud deployment.
Data Science & Forecasting
Data pipelines, analytics workflows, and forecasting systems using Python, SQL, PySpark, SARIMA/LSTM, and stakeholder-ready reporting.
Computer Vision & Medical AI
Research and project experience in thermal imaging, deep learning, image enhancement, face/deepfake detection, and video intelligence pipelines.
Featured work
Selected systems & projects.
AI products, LLM evaluation, data pipelines, and applied computer vision.
Founder · AI Product
Aureum — AI Financial Copilot
Founded and built an AI financial copilot: users manage personal finances and weigh investment decisions through plain conversation. I designed the AI orchestration, financial analytics, market intelligence pipelines, and the cloud infrastructure underneath, and kept every recommendation transparent and non-custodial.
- AI orchestration
- Financial analytics
- Market intelligence
- Cloud infrastructure
GenAI · Computer Vision · Media
Ayclip — AI Video Clipping Pipeline
AI-powered video clipping architecture that turns long-form video into short-form clips: YOLOv11 + MediaPipe + ByteTrack for face tracking and 9:16 auto-reframing, Claude-based semantic highlight detection, Whisper speech-to-text with word-level subtitle synchronization, and asynchronous processing on Docker and AWS ECS/S3.
- Python
- FastAPI
- Celery
- YOLOv11
- MediaPipe
- ByteTrack
- Whisper
- FFmpeg
- Docker
- AWS ECS/S3
LLM Evaluation
Alignerr — LLM Evaluation & Code Review
I evaluate AI-generated STEM, coding, ML, and data science reasoning. In practice that means reviewing code for logic errors and edge cases, writing structured feedback for LLM training workflows, and auditing annotation quality against rubrics.
- Python
- Rubric evaluation
- RLHF/SFT
- QA workflows
RAG · Backend AI
Hospital Chatbot Appointment System
A medical chatbot that took a manual appointment workflow and made it self-service. Under the hood: a LangChain RAG pipeline, document retrieval, multi-intent handling, and automated doctor assignment on a FastAPI backend.
- Python
- FastAPI
- Docker
- LangChain
- RAG
- Caching
AI Assistant · DS Tooling
DeciSense — AI Assistant for Data Science Workflows
Python assistant that automates data science workflows: dataset intake, validation, profiling, task inference, planning, modeling, evaluation, and reporting. Simple inspectable logic first, LLM assistance second.
- Python
- Data validation
- Profiling
- ML planning
- LLM explanations
Data Science · Forecasting
Transportation Demand Forecasting
Forecasting pipelines that join taxi, flight, and weather data on PySpark and GCP. The SARIMA and LSTM models improved prediction accuracy by 15% and cut pipeline runtime by 90%.
- Python
- PySpark
- GCP
- SARIMA
- LSTM
Case-study details for each system are available on request.
Capability matrix
Technical capability, grouped by function.
AI Engineering & LLM Systems
- Agentic AI & multi-agent systems
- MCP (Model Context Protocol)
- RAG · hybrid retrieval · reranking
- Vector databases
- LangChain
- Context & prompt engineering
- Structured outputs & tool calling
- LLM evaluation & LLM-as-judge
- RLHF/SFT evaluation
- Guardrails & AI safety
- Prompt caching & cost optimization
Backend & AI Product Engineering
- Python
- FastAPI
- REST APIs
- SQL
- Caching
- Docker
- Git
- Celery
- Workflow automation
Data Engineering & Analytics
- PySpark
- ETL pipelines
- Batch processing
- Schema design
- BigQuery
- Looker Studio
- Dashboard design
- Reporting automation
Data Science & Machine Learning
- Scikit-learn
- NumPy
- Pandas
- Time-series forecasting
- ARIMA
- ARIMAX
- SARIMA
- SARIMAX
- LSTM
- GRU
- Statistical analysis
- Clustering
- Classification
- Regression
- Feature engineering
- Model evaluation
Computer Vision & Media AI
- OpenCV
- CNN classification
- Medical image analysis
- Thermal imaging
- Deepfake detection
- YOLOv11
- MediaPipe
- ByteTrack
- Whisper
- FFmpeg
Cloud & Deployment
- AWS ECS
- AWS S3
- AWS Bedrock
- AWS Lambda
- CloudFront
- GCP
- BigQuery
- Compute Engine
- Vertex AI
- Streamlit
- Static hosting
AI-Accelerated Engineering
Most of this site, including the recruiter chatbot in the corner, was built pair-programming with AI agents. Working this way is how I ship faster than a one-person team usually can.
- AI coding agents (Claude Code)
- Spec-driven development
- AI code review & auditing
- Eval-driven development
- LLM training data quality
- Human-in-the-loop workflows
- Rapid AI prototyping
Product, Research & Collaboration
- Business problem framing
- Technical documentation
- Stakeholder reporting
- Async collaboration
- Fast iteration
- Startup leadership
- Research writing
Experience
Seven roles in five years.
-
Jan 2026 — Present
Founder · Aureum
Building an AI financial copilot end to end: AI orchestration, financial analytics, market intelligence pipelines, and cloud infrastructure.
-
Mar 2026 — Present
AI Engineer · Ayclip
Designing AI video clipping architecture: CV pipelines, semantic highlight detection, ASR/subtitles, and asynchronous media processing on AWS.
-
Oct 2024 — Present
AI Evaluation Consultant · Alignerr
Evaluating AI-generated STEM/coding/ML solutions, reviewing code quality, and auditing annotation outputs for rubric adherence and reasoning quality.
-
May 2024 — Oct 2025
AI Trainer & LLM Evaluator · Outlier
Reviewed and annotated AI responses for RLHF/SFT across general, technical, and CS domains; assessed instruction-following, truthfulness, and reasoning.
-
Jan 2024 — Jul 2024
Data Scientist Intern · Data Glacier
Built PySpark/GCP pipelines integrating taxi, flight, and weather data; SARIMA & LSTM forecasting improved accuracy 15% and cut runtime 90%.
-
2023
Data Analyst Fellow · GoTo Impact Foundation
Designed dashboards and automated reporting for student performance tracking; translated findings for non-technical stakeholders.
-
Aug 2021 — Feb 2022
AI Engineer Apprentice · Orbit Future Academy
Led a team of four building a deepfake detection system with CNN architectures (MTCNN, InceptionResNetV1), from preprocessing to evaluation.
Research & education
Medical computer vision research.
Master's thesis · Universitas Gadjah Mada
Early detection of diabetic foot ulcers using thermogram and temperature data integration
97.06% reported classification accuracy
Combined plantar thermogram imaging with temperature data to catch diabetic foot ulceration risk earlier. The work covered image enhancement, feature integration, and model evaluation.
Published in SINTECHCOM Journal, Feb 2025 · DOI 10.59190/stc.v5i2.273Education
M.Cs. Computer Science
Universitas Gadjah Mada · GPA 3.79/4.00
Education
S.Kom. Informatics
Universitas Muhammadiyah Malang · GPA 3.86/4.00
Book
Java Itu Mudah
DIVA Press, 2020 · ISBN 9786023918980
Contact
Let's build something practical.
Interested in AI engineering, GenAI, LLM systems, data science, or applied AI product work? Reach me by email or LinkedIn. I'm open to roles and collaborations.
nurilhuda3333@gmail.com