01
Recommendation System
Neural Recommender
A hybrid collaborative-filtering engine blending matrix factorization with transformer-based session embeddings for personalized discovery.
PyTorchFAISSAirflow
→
lailsya
Data Scientist
Data Scientist
Portfolio · 2026
Role
Data Scientist
Based in
Batu, East Java
I design intelligent systems that learn from context — from recommendation engines to graph neural networks — turning messy signals into calm, trustworthy products.
Available
for select collaborations
(About)
Based between the volcanic hills of Batu and the humming edges of Jakarta, I build machine learning systems for teams that care about how their products feel — not just how they score.
My work spans recommendation systems, graph neural networks, forecasting and NLP. I'm happiest when a notebook grows quietly into production, observable, tested, and kind to the people who maintain it.
5+
Years in ML
20
Models shipped
12
Papers read /wk
∞
Cups of kopi
(Selected work)
A curated sample — from research prototypes to production systems.
Recommendation System
A hybrid collaborative-filtering engine blending matrix factorization with transformer-based session embeddings for personalized discovery.
Graph Neural Network
A GNN pipeline for fraud detection on transaction graphs, using GraphSAGE and attention aggregation over billion-edge subgraphs.
Time-series ML
Probabilistic demand forecasting for retail SKUs using Temporal Fusion Transformers, with calibrated prediction intervals.
Language Modeling
Fine-tuned Indonesian language embeddings for sentiment and intent, served on a low-latency ONNX runtime.
(Let's build)
I'm open to selected collaborations, research partnerships, and speaking invitations. Reach out — I usually reply within a day.