Type of position: post-doc
Name: Fitria Wulandari Ramlan
Description:
I completed my PhD in June 2026 on symbolic regression and genetic programming. My thesis was “Symbolic Regression with Genetic Programming: Synthetic Data Generation and Model Selection for Interpolation and Extrapolation.” Most of my work is about making GP models more reliable when they hit data outside their training range: I built a synthetic data method (using KDE plus NN/RF/GP teacher models) that improved extrapolation performance across benchmark datasets, and ran a large comparison of five model-selection criteria (MSE, AIC, BIC, MDL, PSM) across 20 datasets. I also developed new model-selection penalties for extrapolation and interpolation sensitivity, tested on a real-world energy-system model (ETSAP-TIAM), cutting error by about 98% over baseline.
My work has been published at EuroGP (extended version submitted to GPEM by invitation), GECCO, LOD, and ECTA. I currently lead a 7-person research team applying GP surrogate modelling to global energy-system planning, and we’re preparing this work for a journal submission.
Short CV:
PhD in AI (Genetic Programming & Symbolic Regression), University of Galway, 2026. Website: https://www.fitriaramlan.com/
Thesis: Symbolic Regression with Genetic Programming: Synthetic Data Generation and Model Selection for Interpolation and Extrapolation
Research focus:
My work is about making GP models more reliable when they see data unlike what they were trained on in the interpolation/extrapolation problem.
Description:
- Built a synthetic data method using KDE to find sparse spots in the input data, then used NN/RF/GP teacher models to generate extra training data. This gave statistically significant gains in extrapolation performance, and lower error growth as inputs moved further from the training data, across benchmark datasets (published on GECCO 2025).
- Built an automated pipeline comparing five model-selection criteria (MSE, AIC, BIC, MDL, PSM) across 20 benchmark datasets. Found no single criterion works best everywhere, it depends on the dataset and the shape of the Pareto front (published on ECTA 2025).
- Extended standard model-selection metrics with three new penalty terms for extrapolation divergence and interpolation/extrapolation sensitivity. Two resulting configurations (MVP-BIC, MVP-MDL) cut test MSE by about 98% on a real energy-system surrogate model, without extra tuning for that domain (published on EuroGP 2026, and extended version submitted to GPEM through invitation).
- Visiting researcher at HEAL Lab, Austria (2024), working on a cross-institutional comparison of model-selection criteria for symbolic regression (published on ECTA 2025).
Applied work:
Currently leading a 7-person research team on building GP surrogate models for ETSAP-TIAM, a global energy-system model used in climate policy planning. A manuscript on this work is in preparation, with me as first author.
Master’s research (2018–2020):
MSc in Electronics Engineering, Kyungpook National University (GPA 3.88/4.3). Thesis: Evolutionary Multi/Many-objective Approaches for Next Release Optimization Problem. As a research assistant at ECIS Lab, I developed coverage path planning for autonomous greenhouse robots (published IEEE, 2019), applied Differential Evolution to interior design optimisation, and proposed a many-objective evolutionary algorithm using hierarchical Pareto-dominance (published on ICAISC 2019). Three further papers came out of this work, in IEEE Access, IEEE ICTC, and ICAISC (2019).
Publications: IEEE, EuroGP, GECCO, LOD, ECTA (extended paper invited to GPEM), Google scholar: https://scholar.google.com/citations?user=REKn9g0AAAAJ&hl=en
Other: Founded and led a 10-person startup team (web development) for two years before the master’s degree. Teaching assistant and mentor for undergrad/postgrad students during the PhD.
Looking for: Postdoc or research roles in GP, symbolic regression, or model reliability, open to applying these methods in new domains.
Email: rwfitri8@gmail.com
ORCID: 0000-0001-5415-2534
Web page: https://www.fitriaramlan.com
Other link: https://www.linkedin.com/in/fitria-wulandari-ramlan-phd-815918ab
General location: Europe
Specific countries: Austria, Finland, Netherlands, Norway, Sweden, Switzerland, United Kingdom
Until when: 2026-08-01
Present at next GECCO? yes