Available for new roles · Singapore
Marcus Neo
Founding ML Engineer at Datature
4+ years owning the full machine learning lifecycle at an AI startup — from training pipeline architecture and model export to inference serving and data annotation tooling. Built and scaled production systems across classical computer vision, 3D medical imaging, and vision-language model fine-tuning on GCP.
Experience
Founding Machine Learning Engineer (Team Lead)
Datature
Jul 2022 — Present
Model Training Service
Sole author. Framework-agnostic training pipeline supporting JAX, TensorFlow, and PyTorch with multi-GPU support. Onboarded 16+ model architectures across classification, detection, segmentation, pose estimation, anomaly detection, and 3D medical imaging.
3D Medical Imaging Pipeline
End-to-end volumetric training and inference using MONAI (SwinUNETR, SegResNet, nnUNet). NIfTI/DICOM ingestion with RTSTRUCT output for clinical radiotherapy workflows.
VLM Fine-Tuning Pipeline
Built from scratch — LoRA fine-tuning with quantization, tensor parallelism for multi-GPU training. Supports QWEN2.5-VL, NVILA, CosmosReason, KimiVL with VQA, Chain-of-Thought, and video training.
Model Export & Inference
Cross-framework export to TFLite, ONNX, CoreML, TensorRT, OpenVINO with Float16/Int8 quantization. Triton Inference Server integration with Numba JIT optimisation on GKE.
Annotation & Data Services
Import/export for 10+ annotation formats (COCO, LabelMe, SuperAnnotate, ScaleAI). Video tracking with SAM2, VQA annotation pipelines, and Mosaic/CutMix augmentation preview.
CI/CD & DevOps
GitHub Actions pipelines across all ML services — automated linting, testing, Docker builds, and deployment to GCP (Compute Engine, Cloud Run, GKE). Maintained shared PyPI and npm packages.
Machine Learning Intern
Datature
Apr 2021 — Jan 2022
Training & Tooling
Built custom augmentations for TensorFlow training, dataset statistics aggregation, and hyperparameter recommendations. Developed the full backend for Portal — a neural net visualisation webapp with multi-framework support.
Skills
Languages
ML Frameworks
Serving & Export
VLM Fine-Tuning
Cloud & Infrastructure
Education
Imperial College London
MEng Electronic & Information Engineering, First Class Honours
Final Year Project: Data Stream Evolution using Multiresolution Wavelet Transform — ML research applying wavelet transforms for anomaly detection in data streams
Oct 2018 — Jun 2022
Hwa Chong Institution
GCE A-Levels — AAAA
Jan 2014 — Dec 2015
Activities & Interests
Scuba Diving
PADI Advanced Open Water · Nitrox · Drysuit · 50+ Logged Dives
Bali, Indonesia · Ko Tao, Thailand · Malapascua, Philippines · Osezaki, Japan · Tioman, Malaysia
Snowboarding
Picking it up one run at a time
Norikura Kogen, Japan · Yabuli, China
Contract Bridge
Serious hobbyist — always looking for a table
2/1 GF · 15-17 NT · Jacoby 2NT · Bergen Raises · RKCB 1430
Want to learn? Try this! →