National Research Repository
The National Research Fund facilitates research for the advancement of Science, Technology and Innovation. One of our core functions is to compile and maintain a national database of research and innovation projects funded by the Fund and other agencies as per the STI Act of 2013.
Browse Collections
Select a community to browse its collections.
Registry of Repositories in Kenya (RoRiK)
NRF is developing a Registry of Research Repositories in Kenya (RoRiK) in an effort to promote access to research data in the country.
Recent Submissions
Building a Transformer-Based Neural Machine Translation System for English–Kibajuni Translation: A Low-Resource Deep Learning Approach for Indigenous Language Preservation
(South Sahara Academic Publishing South Sahara Multidisciplinary, 2026-07) Anwar H. Ahmed, Wahida M. Bana, Mathew M. Egessa, Collins Kiprotich, Athman L. Omar
Recent progress in artificial intelligence has pushed machine translation to high levels of accuracy
for widely resourced languages. Yet for many indigenous and endangered languages, comparable tools
remain absent, largely because digitized linguistic data are scarce. Kibajuni, a minimally documented
Bantu language spoken along the Kenyan coast, illustrates this gap. Publicly accessible English–Kibajuni
machine translation systems are not available, which restricts both everyday digital use and broader
language preservation work.This paper reports the design, construction, and assessment of a compact
Transformer-based Neural Machine Translation (NMT) system for English–Kibajuni translation.
Training relied on a community-produced parallel corpus of roughly 10,000 aligned sentence pairs. A
Design Science Research approach guided development of the full translation workflow, beginning with
corpus preparation and continuing through Byte Pair Encoding (BPE) tokenization, a custom encoder
decoder Transformer, supervised training in PyTorch, beam-search decoding at inference time, and
deployment as a web application. Because data were limited, emphasis was placed on training stability and
generalization rather than increasing model size. The system therefore integrated AdamW, OneCycle
learning-rate scheduling, dropout, label smoothing, gradient clipping, mixed-precision training, and early
stopping driven by validation BLEU. Results indicate that, despite the small dataset, the model learned
usable semantic correspondences between English and Kibajuni while remaining computationally light.
The final network contains about 6–8 million parameters, occupies roughly 27 MB, and supports real
time translation on modest hardware. In practical terms, the work provides one of the earliest operational
English–Kibajuni neural translation platforms. At the methodological level, it offers a reproducible
template for developing MT systems for other under-resourced African languages. Taken together, the
findings suggest that appropriately scaled Transformer models, paired with subword tokenization and
carefully tuned training procedures, can materially advance digital inclusion and language preservation
for endangered languages.
AI-ENABLED CULTURAL HERITAGE PRESERVATION AND A BROWSER-BASED VIRTUAL MUSEUM: DESIGN AND IMPLEMENTATION OF A KENYA COASTAL HERITAGE PLATFORM
(Open Access Research Journal of Science and Technology,, 2026-09-29) Anwar Hood Ahmed, Mathew Egessa, Patrick Oginga, Collins Becky and Titus Nyandoro
Coastal Kenya holds a wide-ranging cultural heritage expressed in material artefacts, oral traditions, indigenous and
local languages, historical knowledge, and everyday community practices. Yet these resources are increasingly strained
by modernisation and global circulation, shifting modes of cultural transmission, and gaps in systematic documentation.
This study reports the design and implementation of the Kenya Coastal Heritage Digital Platform produced under the
Leveraging Artificial Intelligence for the Preservation and Promotion of Coastal Cultural Heritage in Kenya (LAIPCH-K)
project. The study examines how the project’s initially AI-focused aims were converted into a connected digital heritage
system comprising a public heritage repository, a curator-mediated curation dashboard, and a browser-based virtual
museum. Structured metadata, verified artefact records, multilingual resources, storytelling and oral-history materials,
interactive 3D artefacts, and community-oriented governance arrangements are integrated into a single workflow. The
virtual museum combines Three.js, WebGL, WebXR, Web Audio, and Web Speech, with procedurally generated Swahili
inspired architectural elements, first-person navigation, on-demand loading of validated 3D artefacts, and an animated
multilingual cultural guide named Amina. The study follows a digital-heritage design and implementation pathway
grounded in the LAIPCH-K proposal, project documentation, engineering specifications, and prototype development.
The results demonstrate an operational pipeline linking heritage contribution to curation and validation, metadata
management, publication, and immersive presentation. The evidence does not, however, establish visitor learning
outcomes, community satisfaction, AI-model performance, or broader socioeconomic effects. The study argues that
durable digital heritage preservation depends on aligning technical systems with metadata practice, provenance,
cultural validation, multilingual representation, accessibility, governance, and sustained community participation.
Building a Transformer-Based Neural Machine Translation System for English–Kibajuni Translation: A Low-Resource Deep Learning Approach for Indigenous Language Preservation
(South Sahara Academic Publishing South Sahara Multidisciplinary, 2026-07) Anwar H. Ahmed, Wahida M. Bana, Mathew M. Egessa, Collins Kiprotich, Athman L. Omar
Recent progress in artificial intelligence has pushed machine translation to high levels of accuracy
for widely resourced languages. Yet for many indigenous and endangered languages, comparable tools
remain absent, largely because digitized linguistic data are scarce. Kibajuni, a minimally documented
Bantu language spoken along the Kenyan coast, illustrates this gap. Publicly accessible English–Kibajuni
machine translation systems are not available, which restricts both everyday digital use and broader
language preservation work.This paper reports the design, construction, and assessment of a compact
Transformer-based Neural Machine Translation (NMT) system for English–Kibajuni translation.
Training relied on a community-produced parallel corpus of roughly 10,000 aligned sentence pairs. A
Design Science Research approach guided development of the full translation workflow, beginning with
corpus preparation and continuing through Byte Pair Encoding (BPE) tokenization, a custom encoder
decoder Transformer, supervised training in PyTorch, beam-search decoding at inference time, and
deployment as a web application. Because data were limited, emphasis was placed on training stability and
generalization rather than increasing model size. The system therefore integrated AdamW, OneCycle
learning-rate scheduling, dropout, label smoothing, gradient clipping, mixed-precision training, and early
stopping driven by validation BLEU. Results indicate that, despite the small dataset, the model learned
usable semantic correspondences between English and Kibajuni while remaining computationally light.
The final network contains about 6–8 million parameters, occupies roughly 27 MB, and supports real
time translation on modest hardware. In practical terms, the work provides one of the earliest operational
English–Kibajuni neural translation platforms. At the methodological level, it offers a reproducible
template for developing MT systems for other under-resourced African languages. Taken together, the
findings suggest that appropriately scaled Transformer models, paired with subword tokenization and
carefully tuned training procedures, can materially advance digital inclusion and language preservation
for endangered languages.
Governance Practice Implementation on Education for Sustainable Development in Host Universities’ Regional Centres of Expertise, Kenya
(Journal of Education, Society and Behavioural Science, 2021-09-10) Nancy Ng’ang’a, Ruth Thinguri and Ayub Macharia
Institutional management refers to the effective running of different programmes and activities of an
institution. This study sought to assess governance practice implementation on education for
sustainable development (ESD) in host universities’ Regional Centres of Expertise (RCE) in Kenya.
The study employed mixed methodology and concurrent model of triangulation and descriptive
survey research design to collect both qualitative and quantitative data. The target population was
from 8 acknowledged RCEs in Kenya. This included 8 RCE coordinators, 270 lecturers conversant
with ESD and 150 representatives of partner institutions making a total of 428 respondents.
Purposive and simple random sampling was used to come up with the study sample of 93
respondents comprising of of 3 RCE coordinators, 60 lecturers, and 30 representatives of partner
institutions. Research tools used were open and close-ended questionnaires, interview schedule
and document analysis checklist. Quantitative data was analysed descriptively using SPSS version
23 and presented using tables, frequencies, percentages and Pearson’s correlation coefficient.
Qualitative data was thematically analysed and presented verbatim and as narratives using MS
Word. Mixing and interpretation of data was then done. The study established that despite
numerous challenges, RCE host universities had implemented RCE governance practice on ESD
through establishment of governance structures and appointment of coordinators. Financial and
infrastructural resources were found to be a main challenge in implementation of governance
practice on ESD. Anticipated beneficiaries of the study findings and conclusion include the
management of RCE host universities, RCE coordinators, Ministry of education and County
governments. The study recommended that host universities in Regional Centres of Expertise
should initiate innovative methods of raising funds from both government and private sectors to
establish a funding mechanism for ESD activities and programs in the RCEs. The appointment of
RCE coordinators needs to be removed from the host university Vice Chancellors and a committee
of partner representatives tasked with the appointment.
Analysis of Microfinance Institution Outreach on the Realization of Economic Pillar of Vision 2030: A Case of Kakamega County
(Research Journal of Finance and Accounting, 2019-08-31) Sangoro Oscar, Gongera George, Ronald Onsiro Ronald
This paper sought the analysis of microfinance institutions outreach on the realization of economic pillar of vision
2030.Microfinance institution outreach in this study is checked in terms of breadth and depth. In line with
economic pillars for vision 2030, MFIs are vital in advancing credit to the communities which are financially
constrained but have feasible, practicable and promising investment business ideas. This study was carried in
Kakamega County which has very high level of poverty. This study adopted Cross-sectional research design and
correlation design. The target population in this study consist of businesses deriving their capital from 15 microfinance institutions both small scale and medium scale, Employees of MFIs comprising of branch manager, credit
officer, risk officers and operational manager. The results indicated that MFI outreach had a statistically significant
influence on the realization of vision 2030 in Kakamega. The study concluded that MFI outreach had significant
positive influence on the realization of economic pillar of vision 2030
Keywords: Microfinance institution, Outreach, Vision 2030, Economic Pillar
