A self-grounding concept bottleneck that discovers temporal segments and human-inspectable concepts, with a built-in trust score for unfamiliar inputs. Across 17 benchmarks, it reaches 97.91% accuracy on Epilepsy using only 1% labeled data.
Sachith Abeywickrama
Efficient, trustworthy AI for sequential data.
Ph.D. researcher at NTU and A*STAR working on foundation models, continual learning, and interpretable systems under real-world shift.
Selected research
Ph.D. work under review
Methods for adaptive, interpretable, and data-efficient modeling of non-stationary sequences.
Entropy-guided dynamic patch segmentation that places boundaries at natural temporal changes instead of fixed windows, reducing average MSE by 9.7% against PatchTST across six forecasting benchmarks.
Identity-preserving domain-incremental adaptation for frozen time-series foundation models. With a frozen Sundial backbone, PACT leads all five evaluated streams, including 0.958 average macro-F1 on four bearing conditions, while keeping relative forgetting below 0.05.
Field inference for unknown binary protocols without priors or executables. A pretrained byte model and cross-message transformer achieve 0.807 average boundary F1 and 0.815 recall across 12 held-out protocols, lifting recall from 0.538 to 0.815 over the strongest trace-only baseline.
Path
Research and industry experience
From production LLM systems to distributed foundation-model research.
-
Aug 2024 – Present
Ph.D. Researcher / Research Scholar
Institute for Infocomm Research, A*STAR · Singapore
Building efficient, trustworthy, and continual-learning methods for foundation models under distribution shift and limited labels. Running reproducible PyTorch pipelines and multi-GPU training on SLURM for time-series foundation models and open-source LLMs.
-
May 2023 – Jun 2024
Associate Machine Learning Engineer
Rootcode · Sri Lanka
Delivered document QA, semantic search, NER, and open-source LLM applications for enterprise access problems. Productionized models with FastAPI, Docker, evaluation workflows, and cloud infrastructure.
-
Nov 2022 – Apr 2023
Artificial Intelligence Engineer Intern
Rootcode · Sri Lanka
Built conversational AI and semantic-search systems for e-commerce discovery and support, deploying containerized FastAPI services on AWS with cloud NoSQL backends.
-
Jun 2022 – Sep 2022
Research Intern
Nanyang Technological University · Remote
Worked on data-efficient polymer property prediction using topological molecular features, training convolutional, transformer, and ensemble models with persistent spectral fingerprints.
Background
Education and honors
Education
Nanyang Technological University
Ph.D. Student, Electrical and Electronic Engineering
Singapore · Aug 2024 – Present
Research with A*STAR on efficient, interpretable, and continually adaptive AI for non-stationary sequential data.
University of Colombo
B.Sc. (Hons), Industrial Statistics and Mathematical Finance
Sri Lanka · Jan 2019 – Apr 2023
Honors
Second Runner-up
China-ASEAN Student Innovation & Entrepreneurship Challenge
Guiyang, China · July 2026
Awarded for “Precision Drive Coming,” an autonomous humanoid robotics joint solution representing NTU. Selected among 32 finalists from 1,007 entries across 149 universities.
Singapore International Graduate Award (SINGA)
A*STAR
2024
Competitive Ph.D. scholarship supporting doctoral study at NTU and research at A*STAR’s Institute for Infocomm Research.
Toolkit
Technical skills
Machine learning
PyTorch, TensorFlow, scikit-learn, Transformers, foundation models, LLMs, self-supervised learning, continual learning, generative AI
Trustworthy & efficient AI
Interpretable AI, out-of-distribution detection, few-shot learning, model adaptation, efficient sequence modeling, uncertainty estimation
Applications
Time-series modeling, NLP, semantic search, information retrieval, document QA, named-entity recognition, anomaly detection
Systems
Python, SLURM, distributed multi-GPU training, Docker, Git, FastAPI, AWS, Google Cloud, Microsoft Azure
Connect
Let’s talk research or collaboration.
Based in Singapore. Open to research discussions, collaborations, and academic opportunities.