Akwasi Asare
Nana Akwasi Asare
AI Research Engineer
Software Engineer
Khaya AI GhanaNLP Bagci Lab — Northwestern Univ.
⚡ Latest Updates
New New Paper Published: TransUNet-GradCAM:a hybrid transformer-U-Net with self-attention and explainable visualizations for foot ulcer segmentation at BMC Medical Imaging Journal (Q1).
New Selected as a 2026 Africa Healthcare Innovation Fellow (AHIF) by InnovateHealth Africa — cohort of 45 fellows from 3,500+ applicants.
New Preprint: Nsanku — evaluating zero-shot translation of 19 LLMs across 43 Ghanaian languages. Submitted May 2026.
New Preprint: Ayoo — multilingual multimodal MIR dataset for Ghana Music Award videos. DOI: 10.21203/rs.3.rs-9475824/v1
New Polyp segmentation paper published at ACDSA IEEE Conference, Boracay, Philippines, 2026.
New Published: GhanaNLP Parallel Corpora — 41,513 sentence pairs across 5 Ghanaian languages. arXiv:2603.13793.
New Invited speaker at Women and AI event hosted by Tech Labari at Google AI Community Center, Accra.
New Panelist & Research Lead at UNICEF AI & NLP Summit, Dakar, Senegal — representing Khaya AI & GhanaNLP.
Published: MonkeyPox Detection & Anemia Detection papers in Springer CCIS (AIR-RES 2025), January 2026.
Currently Research Trainee at Bagci Lab, Northwestern University (Machine and Hybrid Intelligence Lab).
Attended PRIMED-AI Workshop at NIH (2025) on precision medicine AI.
AI Researcher & Engineer

Building AI that speaks
every language.

I am an AI Research Engineer focused on deep learning, computer vision, NLP, and software engineering for healthcare, agriculture, and low-resource African languages. My work spans AI-powered medical imaging, multilingual clinical NLP, agro-computing, and the application of rigorous software engineering principles to AI systems — with the mission that no African is left behind in the digital age.

"Adding more layers and adjusting the hyperparameters for victory and glory." — Akwasi Asare
Research Interests
  • › Computer Vision & Medical Imaging Computing
  • › NLP for Healthcare & African Languages
  • › Agro-Computing & Precision Agriculture
  • › Software Engineering for AI Systems
  • › Explainable AI (XAI)
Education
B.Sc. Software Engineering
Ghana Communication Technology University · 2021–2024
A.S. Health Sciences
University of the People, USA · 2020–2024

Experience

AI Research Engineer

Khaya AI Company · Sept 2025 – Present

Machine Learning Researcher

GhanaNLP · Sept 2024 – Present

AI Research Assistant

Machine and Hybrid Intelligence Lab (Bagci Lab), Northwestern University · May 2025 – Present

Co-Founder & Deputy Director

Centre for Medical and Agro Computing (C-MAC)

Research Output

Selected Publications

00
Performance Evaluation DR

TransUNet-GradCAM: a hybrid transformer-U-Net with self-attention and explainable visualizations for foot ulcer segmentation BMC Medical Imaging

Akwasi Asare,Mary Sagoe, Justice Williams Asare, Stephen Edward Moore

BMC Medical Imaging Journal, 2026

Automated segmentation of diabetic foot ulcers (DFUs) supports clinical diagnosis, treatment planning, and longitudinal wound monitoring, but remains difficult owing to the heterogeneous appearance, irregular morphology, and cluttered backgrounds of ulcers in clinical photographs.

Journal
2026
01
Performance Evaluation DR

Performance of ResNet-18 and InceptionResNetV2 in Automated Detection of Diabetic Retinopathy Medicine Advances

Akwasi Asare, Alvin Adjei Broni, Alex Kwasi Asare Dickson, Mary Sagoe, Joshua Makafui Cudjoe

Health Informatics Journal (Medicine Advances), 2025, 3(3): 231–241

Benchmarks two deep learning architectures — ResNet-18 and InceptionResNetV2 — for automated diabetic retinopathy grading, identifying which model delivers more reliable clinical performance for low-resource ophthalmic screening.

Journal
2025
02
Polyp Segmentation

Improved Segmentation of Polyps and Visual Explainability Analysis IEEE ACDSA 2026

Akwasi Asare, Thanh-Huy Nguyen, Ulas Bagci

2026 Intl. Conference on AI, Computer, Data Sciences and Applications (ACDSA), Boracay Island, Philippines, pp. 1–8

Proposes an enhanced U-Net model with ResNet-34 encoder for colorectal polyp segmentation, and applies Grad-CAM visualizations to explain model decisions — improving both accuracy and clinical trustworthiness for gastrointestinal cancer screening.

IEEE DOI Presentation
2026
03
Monkeypox Detection

Enhanced Ensemble Model-Based Approach for Monkeypox Detection Using Medical Images Springer CCIS

Akwasi Asare, Justice Williams Asare, Martin Mabeifam Ujakpa, Emmanuel Akwah Kyei, Laizah Sashah Mutasa, Emmanuel Freeman, William Leslie Brown-Acquaye, Forgor Lempogo

AI Revolution: Research, Ethics and Society (AIR-RES 2025), Springer CCIS Vol. 2723, pp. 166–189

Develops an ensemble deep learning model combining multiple CNN architectures to improve monkeypox skin lesion detection from medical images, addressing the urgent need for fast, scalable diagnostic tools during viral outbreaks in Africa.

Springer
2025
04
Anemia Detection

A Hybrid Three-Layer CNN Architecture for Detecting Anemia in Children Using Palpebral Conjunctiva Images Springer CCIS

Justice Williams Asare, Akwasi Asare, Martin Mabeifam Ujakpa, Emmanuel Akwah Kyei, Laizah Sashah Mutasa, William Leslie Brown-Acquaye, Forgor Lempogo, Alfred Coleman

AI Revolution: Research, Ethics and Society (AIR-RES 2025), Springer CCIS Vol. 2723, pp. 144–156

Introduces a non-invasive deep learning approach that detects anemia in children by analyzing inner eyelid (palpebral conjunctiva) images using a hybrid three-layer CNN — offering a low-cost alternative to blood-based tests in resource-limited settings.

Springer
2025
05
Rabies Outbreak Prediction

Machine Learning-Based Prediction of Rabies Outbreaks Using Epidemiological and Environmental Data in Africa Springer

Akwasi Asare, Justice Williams Asare, Seth Alornyo, Emmanuel Akwah Kyei, Emmanuel Freeman, William Leslie Brown-Acquaye, Martin Mabeifam Ujakpa, Israel Edem Agbehadji, Richard C. Millham

Intl. Conference on Computational Science and Computational Intelligence, Springer Nature, 2024, pp. 153–167

Applies machine learning models to epidemiological and environmental datasets to predict rabies outbreak hotspots across Africa, enabling public health authorities to prioritize intervention resources and reduce transmission risk.

Springer
2024
06
Nsanku LLM Translation

Nsanku: Evaluating Zero-Shot Translation Performance of LLMs for Ghanaian Languages Preprint

Stephen E. Moore, Mich-Seth Owusu, Akwasi Asare, Lawrence Adu Gyamfi, Paul Azunre, Joel Budu, Jonathan Asiamah, Elias Dzobo, Kelvin Newman, Edmund O. Benefo, Gerhardt Datsomor, Onesimus Addo Appiah, Ama Branoa Banful, Lucas Woedem Kpatah, Saani Mustapha Deishini, John Ayernor

arXiv · Submitted May 2026

The most comprehensive evaluation of LLM translation to date for African languages — benchmarks 19 open-weight and proprietary LLMs on zero-shot translation across 43 Ghanaian languages, revealing that no current model achieves reliable, consistent performance for Ghanaian language translation at scale.

arXiv
2026
07
Ayoo Music Dataset

Ayoo: A Multilingual Multimodal Music Information Retrieval Dataset for Ghana Music Award Videos Preprint

Stephen E. Moore, Akwasi Asare, Samuel Kwadwo Kubiti

Research Square Preprint, 2026 · Univ. of Cape Coast · Ghana NLP · Center for AI Research, Ghana

Introduces Ayoo, a first-of-its-kind multilingual and multimodal dataset built from Ghana Music Award videos, enabling research in music information retrieval, audio-visual understanding, and NLP for Ghanaian languages including Twi, Ga, and Ewe.

DOI
2026
08
GhanaNLP Parallel Corpora

GhanaNLP Parallel Corpora: Comprehensive Multilingual Resources for Low-Resource Ghanaian Languages Preprint

Lawrence Adu Gyamfi, Paul Azunre, Stephen Edward Moore, Joel Budu, Akwasi Asare, Mich-Seth Owusu, Jonathan Ofori Asiamah

arXiv:2603.13793, 2026

Presents 41,513 human-curated parallel sentence pairs across Twi, Fante, Ewe, Ga, and Kusaal — providing the foundational multilingual data infrastructure that powers the Khaya AI translation engine and supports machine translation research for underrepresented Ghanaian languages.

arXiv
2026
09
TransUNet-GradCAM Foot Ulcer

TransUNet-GradCAM: A Hybrid Transformer-U-Net with Self-Attention and Explainable Visualizations for Foot Ulcer Segmentation Preprint

Akwasi Asare, Mary Sagoe, Justice Williams Asare, Stephen Edward Moore

arXiv:2508.03758, 2025

Combines a Transformer-based U-Net with Grad-CAM explainability to accurately segment diabetic foot ulcers — helping clinicians visualize what the AI focuses on, improving trust and enabling better wound management decisions in clinical practice.

arXiv
2025
10
PerceptronCARE

PerceptronCARE: A Deep Learning-Based Intelligent Teleophthalmology Application for Diabetic Retinopathy Diagnosis Preprint

Akwasi Asare, Isaac Baffour Senkyire, Emmanuel Freeman, Mary Sagoe, Simon Hilary Ayinedenaba Aluze-Ele, Kelvin Kwao

arXiv:2509.18160, 2025

Presents PerceptronCARE, a mobile teleophthalmology system that uses deep learning to diagnose diabetic retinopathy remotely — designed to extend specialist-level eye care to patients in underserved communities where ophthalmologists are scarce.

arXiv PDF
2025
11
Edge DR Deployment

Deploying and Evaluating Multiple Deep Learning Models on Edge Devices for Diabetic Retinopathy Detection Preprint

Akwasi Asare, Dennis Agyemanh Nana Gookyi, Derrick Boateng, Fortunatus Aabangbio Wulnye

arXiv:2506.14834, 2025

Evaluates the real-world feasibility of running multiple deep learning models directly on low-power edge devices for diabetic retinopathy detection — comparing accuracy, latency, and efficiency trade-offs to guide deployment in off-grid and resource-constrained clinical environments.

arXiv PDF
2025
12
Ndwom Music Dataset

Ndwom: A Multimodal Music Information Retrieval Dataset for Akan Musical Videos Preprint

Stephen Edward Moore, Nana Akwasi Asare, Samuel Kwadwo Kubiti

Research Square Preprint, 2024

Introduces Ndwom, the first multimodal music information retrieval dataset built from Akan-language music videos — bridging a critical gap in African audio-visual NLP research and enabling sentiment analysis, genre classification, and music recommendation in the Akan language.

DOI
2024
Engagement

Talks, Fellowships & Events

<AHIF 2026 Fellowship
2026 Fellowship

Africa Healthcare Innovation Fellow — AHIF 2026

Selected from 3,500+ applicants as one of 45 fellows for the Africa Healthcare Innovation Fellowship (AHIF) by InnovateHealth Africa. The 12-week programme equips professionals to lead healthcare transformation across Africa. Focus: scalable, context-aware AI health solutions.
→ Announcement post

UNICEF AI & NLP Summit Dakar
Panelist & Research Lead

UNICEF AI & NLP Summit — Dakar, Senegal

Represented Khaya AI and GhanaNLP at the UNICEF AI & NLP Summit as both panelist and Research & Implementation Lead on ASRH. Announced scaling from Ghanaian to 32+ African languages serving 500M+ people, and partnership with UNICEF to develop ASRH datasets in local languages.
→ LinkedIn post

Women and AI, Google AI Community Center Accra
Invited Speaker

Women and AI — Google AI Community Center, Accra

Invited speaker at the Women and AI event hosted by Tech Labari at the Google AI Community Center in Accra, representing Khaya AI. Talk: "Linguistic Justice: What Women in Tech Need to Know About Language, Power and AI." Discussed how biases in language and training data shape AI systems and affect representation, fairness, and opportunity.
→ Read event coverage — Tech Labari

Talks & Blogs
CMAC

Facilitator: Data Science for Medical Imaging & Agro Computing

Lead speaker at the Center for Medical and Agro Computing (CMAC) Mentorship Program, Koforidua Technical University, 2025

CSCI

Rabies Outbreak Prediction using Machine Learning

Intl. Conference on Computational Science and Computational Intelligence (CSCI), Las Vegas, 2024

Medium

Bridging the Gap: Why Africa Lags Behind in the AI Revolution

Medium blog post examining structural, economic, and data challenges holding back African AI adoption

Workshops & Programs
GE4H

Climate Change, Breast Cancer & AI: Intersections, Impacts, and Solutions

Speaker at Green Earth for HUMANITY (GE4H) webinar on AI for breast cancer detection and climate-health intersections. Oct 2025

GIZ

AI Practitioners' Guide Development for Ghana (GIZ / BMZ Initiative)

Invited participant developing Ghana's AI ethics guide — FAIR Forward program, implemented by GIZ, led by Heritors Labs. June 2025

SPARK

SPARK Academy — AI for African Medical Imaging

Sprint AI Training for African Medical Imaging Knowledge Translation (CAMERA Consortium). Focus: brain tumor segmentation, 2024

Ongoing Work

Active Projects

👁️

EyeGPT — DR Detection

LLM-based AI system for Diabetic Retinopathy detection combining vision models and language intelligence for clinical decision support in resource-constrained settings.

Vision-Language Medical Imaging Deployment
⌚

Wearable & Edge Device for Diabetes Intervention

AI-powered diabetic detection and monitoring device designed for deployment in low-resource settings across Africa.

Edge AI Mobile Health IoT
🦷

MICCAI STSR 2025 — Teeth Root Canal Segmentation

Semi-supervised segmentation of tooth root pulp canals in 3D CBCT scans using UNet with attention mechanisms to reduce annotation burden.

Semi-supervised 3D CBCT Attention UNet
🫀

Enhanced HaD-Net for Pancreas CT Segmentation

Enhanced HaD-Net architecture on a 2D sliced dataset for precise pancreas segmentation in CT imaging.

CT Segmentation HaD-Net 2D Sliced
🎵

Multimodal Akan & Ghanaian Music Recommendation

Development of the Ndwom and Ayoo datasets and sentiment-aware music recommendation systems integrating audio and lyrical metadata across Ghanaian languages.

Music IR Multimodal Low-resource NLP