One PlatformCollaborationsIn the PressFull Output

The science behind
Logic Byte.

Logic Byte's platform is built on a foundation of peer-reviewed research — in machine learning for image-based classification, computational analysis of complex imaging data, and domain-aware pipeline development.

Validated in the lab and hardened through direct engineering experience across two demanding imaging domains: biomedical analytics and semiconductor precision manufacturing.

Research-validated capabilities, deployed in production. For the complete record of publications, collaborations, and press coverage, see the full research output.

20+
Peer-reviewed publications
1,091+
Citations
10+
Years of experience
15+
Imaging modalities

One Platform. Multiple Domains.

The technical depth behind Logic Byte spans both peer-reviewed academic research and direct professional engineering experience — across biomedical analytics, semiconductor precision manufacturing, and everything in-between.

Biomedical & Life Sciences

From imaging research
to diagnostic insight

Peer-reviewed research spanning AI-driven single-cell analysis, nuclear morphology as a diagnostic readout, and label-free in-silico staining (the science behind CytoByte) — the scientific foundation for OptiFlow's biomedical capabilities.

Featured · APL Bioengineering, 2021
Machine learning for single-cell classification by intracellular pH
AI-driven identification of cancer cells from label-free imaging of intracellular acidity — without fluorescent markers. 95% accuracy in under 35 minutes. Covered by The Straits Times, CNA, and international press.
Read paper →
Vision Generative AI · Label-free Imaging · In preparation
In-silico staining from DIC microscopy
Vision generative AI that renders pseudo-DAPI nuclear staining directly from DIC images — no chemical preparation, no phototoxicity. Extends to Phase Contrast and Transmitted Light inputs. Deployed commercially at ETH Zurich. Manuscript in preparation.
AI-driven Diagnostics · NUS Biomedical Engineering
Image-based biosensor for broad-spectrum disease classification
Deep CNN trained on morphological and textural cell features to predict disease state on microfluidic platforms. Physics-aware defocus augmentation lifted accuracy from 92% to 97% — an early demonstration of domain-informed model design.
AI Diagnostics · Communications Biology, 2021
Chromatin biomarkers for tumour progression
Single-cell imaging pipeline that classifies breast tissue biopsies by cancer stage using nuclear and chromatin features — demonstrating AI-based grading at clinical resolution.
Read paper →
Semiconductor & Industrial

From raw image data
to production decisions

Direct professional engineering experience building and deploying inspection and analytics pipelines on commercial electron and optical microscopy instruments — where accuracy and throughput are equally non-negotiable.

Featured · Autonomous Systems
Intelligent target acquisition with automated selective high-resolution imaging
Autonomous imaging system that rapidly identifies regions of interest at low resolution and selectively acquires high-resolution images only at those locations — eliminating the throughput ceiling of full-field high-res scanning while preserving analytical depth where it matters.
Inspection · SEM / Optical
Wafer defect detection & non-destructive testing
End-to-end defect detection capabilities on SEM — supervised classification of known defect classes and unsupervised texture analysis and anomaly detection for novel failure modes. Deployed on a commercial production instrument.
Autonomous Control · Hardware-Software Co-design
Signal processing-based auto-tuning of imaging parameters
Cross-functional hardware-software project developing automated instrument calibration to ensure optimum imaging conditions — removing manual tuning from the workflow.
Performance · Inference Optimisation
Latency reduction · Real-time processing
On-device inference latency reduction through quantisation, weight sharing and cross-attention architecture design. Separately, low-footprint plugins for live-stream processing in embedded systems.
Architecture · Multi-source Fusion
Multimodal AI for heterogeneous imaging sources
Multi-source fusion architectures combining data from heterogeneous imaging modalities into unified analytical outputs — enabling richer inspection decisions from a single pipeline pass.
Selected collaborations

Selected collaborations

Projects undertaken in collaboration with research institutions and industry partners, spanning both domains. Outcomes inform OptiFlow's capabilities directly.

Industry & Commercial
Autonomous Systems 2025
Reinforcement learning for iterative multi-parameter auto-tuning of an imaging device
Embedded Systems 2024
Framework for real-time low computational processing of livestreams
Semiconductor · NDT 2023
Computer vision-based wafer inspection and defect detection
Vision Generative AI · ETH Zurich 2022
Generative AI for augmented imaging — academic collaboration converted to commercial deployment
Media coverage

In the press

Logic Byte's founding research in AI-driven biomedical imaging has been covered internationally — from Singapore's flagship broadcast and print media to specialist AI publications. The full press record is on the research output page.

Channel News Asia Singapore 2019
Becoming Human, Episode 2: Unnatural Genius
↗
The Straits Times Singapore 2021
Researchers use AI to identify the pH of cancer cells
↗
LabelFlow · AI Stories France 2022
Diagnosing breast cancer in 35 minutes instead of days
↗