Scalable AI
Production-grade machine learning pipelines that balance compute efficiency with predictive accuracy, from training and evaluation to serving and monitoring.
Software Engineer · ML ResearcherMSc Computer Science, University of GhanaAccra
Bless
Ziamah
I build production ML systems for fintech and identity, and research robust deep learning for healthcare at DISAL.
98.83%
ViT-B/16 accuracy on chest CT · Springer, 2025
24h → 3m
Onboarding time after ML verification shipped
+85%
Verification accuracy, anti-spoof e-KYC
−40%
API calls after a caching redesign
01
Focus
Systems Architecture
Modular backends designed for resilience and maintainability: payments, trading and verification systems that hold up at volume.
ML for Healthcare
Medical imaging and clinical time-series models that stay reliable under distribution shift, noise and missing data.
02
AllPublications & interestsResearch
SARS detection in chest CT scan images using the bootstrapped ViT-B/16 model
Iran Journal of Computer Science · Springer
A fine-tuned ViT-B/16 reaches 98.83% accuracy on 2,482 chest CT images, outperforming DenseNet201 and VGG19.
Read the paperEarly sepsis prediction from clinical time series
DISAL, University of Ghana · Advisor: Prof. Justice Kwame Appati
Re-encodes physiological signals as Gramian Angular Field images so CNNs can learn pairwise temporal relationships across whole time windows.
About the thesis03
Swipe →Selected work
Anti-Spoof e-KYC System
End-to-end identity verification with built-in spoof detection, improving verification accuracy by 85% for financial services.
View on GitHubCustomer Onboarding ML Engine
Automated ML verification workflows for new users, cutting onboarding from 24 hours to 3 minutes.
Corporate projectAI Grant Search & Application
A platform that automates grant discovery and applications with AI-powered matching.
Visit siteAnti-Spoof e-KYC System
End-to-end identity verification with built-in spoof detection, improving verification accuracy by 85% for financial services.
TensorFlow · OpenCV · Deep Learning
View on GitHubCustomer Onboarding ML Engine
Automated ML verification workflows for new users, cutting onboarding from 24 hours to 3 minutes.
Rails · OpenCV · Face Recognition
Corporate projectAI Grant Search & Application
A platform that automates grant discovery and applications with AI-powered matching.
Node.js · React · Automation
Visit site04
Full historyExperience
2025 – Present
Software Engineer
Black Star Group · Onsite
Self-hosted LLM infrastructure, KYC/AML verification backends, and payment and trading integrations for fintech clients.
2024 – Present
Graduate Research Assistant
DISAL, University of Ghana
Thesis research on early sepsis prediction; published work on vision transformers for medical imaging.
2024 – 2025
Software Engineer
Achieve by Petra · Remote
ML-powered identity verification in production (24h → under 3 min) and Rails services automating customer care.
2024 – 2025
Software Engineer, Part-time
Techvision Global (Accede Ltd) · Remote
Next.js and TypeScript interfaces; a caching redesign that cut API calls 40% and saved $400+ a month.
2024
Junior Machine Learning Engineer
StacAI · Remote
End-to-end e-KYC pipeline and anti-spoofing models, improving verification accuracy by 85%.
05
Stack
Machine learning
- PyTorch
- TensorFlow
- scikit-learn
- Vision Transformers
- CNNs
- Computer Vision
- Anti-spoofing
- OpenCV
- Model deployment
Languages & frameworks
- Python
- Django
- FastAPI
- Ruby on Rails
- NestJS
- TypeScript
- Next.js
- React
- Pandas
- NumPy
- SQL
Infrastructure & tools
- AWS
- GCP
- Azure
- Docker
- PostgreSQL
- MongoDB
- Linux
- Git
- CI/CD