University of Kinshasa researcher (h-index 33) moving into governed LLM systems and efficient language models.
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
33
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
Not yet scored: the output axis would measure publication volume and productivity against field norms, but the underlying data is insufficient here. Yearly counts and 135 recent works exist, yet no normalized output score is available.
Impact—
Not yet scored: the impact axis would measure citation influence and field-weighted citation impact against peers. Despite an h-index of 33 and i10-index of 86 on record, the normalized impact data needed to score this axis is insufficient.
Focus0/100
The focus score of 0 signals maximal dispersion rather than absence of work: the record spreads across at least five distinct topic clusters — localization technologies (35 works), chaos control and synchronization (33), neural networks and applications (30), advanced neural network applications (29), and traffic control and management (25) — while the recent field distribution splits evenly (0.5/0.5) between AI governance / trustworthy LLM systems and efficient language models. This is a deliberately multi-domain trajectory.
Collaboration36/100
The collaboration score of 36 rests on a small verified set: 5 distinct coauthors in the record, with Kabeh Mohsenzadegan and Vahid Tavakkoli appearing most frequently. Both Evolve-reviewed 2026 papers place this researcher in the last author position without corresponding-author status, a pattern consistent with senior or supervisory authorship on small teams. The low score reflects concentration in a compact circle rather than an absence of partners.
Activity100/100
The activity score of 100 reflects an accelerating late-career output curve: yearly counts climb from 32 works in 2024 and 39 in 2025 to 64 already indexed in 2026, with 135 works across the most recent three-year window. This extends a publishing history that begins in 1998 (3 works) and includes earlier peaks of 24 in 2006 and 25 in 2009, making the current surge the highest sustained phase on record.
Reproducibility78/100
The reproducibility score of 78 is the mean across 2 reviewed papers. Policy-as-Skill reaches high reproducibility readiness, packaging governance as versioned executable skills, though Evolve notes its headline gains ride a degenerate subset, a synthetic corpus, and single-expert labels. TinyCeNN-LM scores medium: its quality-gated attention-to-recurrent conversion with accept/rollback is honestly reported, but evidence is small-scale, with no speedup and only marginal cache gains.
Research summary
Kyandoghere Kyamakya is a researcher at the University of Kinshasa (DR Congo) whose bibliometric record, name-resolved from OpenAlex, spans nearly three decades of publishing: yearly output counts begin in 1998 (3 works) and run through a strong recent surge, with 32 works in 2024, 39 in 2025, and 64 already indexed in 2026, and 135 works over the most recent three-year window. Cumulative citation-based indicators are substantial, with an h-index of 33 and an i10-index of 86. The topic profile is unusually broad — the radar focus score is 0 — led by indoor and outdoor localization technologies (35 works), chaos control and synchronization (33), neural networks and applications (30), advanced neural network applications (29), and traffic control and management (25). The two Evolve-reviewed papers, both 2026 arXiv preprints with the researcher in the last author position, sit in newer territory: Policy-as-Skill on governed LLM decision support (reproducibility readiness high) and TinyCeNN-LM on quality-gated attention-to-recurrent conversion (medium). Across those two papers the mean reproducibility score is 78, and the recent field distribution splits evenly (0.5/0.5) between AI governance / trustworthy LLM systems and efficient language models. Collaboration is concentrated: only 5 distinct coauthors appear in the record, with Kabeh Mohsenzadegan and Vahid Tavakkoli the most frequent partners, yielding a radar collaboration score of 36.
Collaboration pattern
The collaboration footprint is narrow: 5 distinct coauthors appear in the raw record, and the recurring partners are Kabeh Mohsenzadegan and Vahid Tavakkoli. Both 2026 Evolve-reviewed papers list Kyamakya in the last author position without corresponding-author flag, consistent with a senior-author role on compact teams rather than large distributed consortia. The radar collaboration score of 36 reflects this concentrated pattern; expanding beyond this circle would raise the score.
Research keywords
Highlights
Governance packaged as executable skills (Policy-as-Skill, 2026)
Quality-gated attention-to-recurrent conversion (TinyCeNN-LM, 2026)
Record output surge in 2024–2026
Bibliometric counts rise to 32 works in 2024, 39 in 2025, and 64 in 2026, totalling 135 works over the most recent three-year window — the most productive phase of a publishing record reaching back to 1998.
Timeline
- 1998First indexed works appear (3 in 1998), marking the start of a publishing record that runs to the present.
- 2006Annual output reaches an early peak of 24 works, within a sustained 2006–2011 run of 13–25 works per year (25 in 2009).
- 2024Output accelerates sharply: 32 works in 2024, after more than a decade in which annual counts stayed between 5 and 22.
- 2026Two Evolve-reviewed arXiv preprints appear — Policy-as-Skill and TinyCeNN-LM, both with the researcher as last author — while 64 works are indexed this year, the highest annual count on record.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric metrics are name-resolved from OpenAlex only; homonym or entity-splitting effects cannot be ruled out.
- The output and impact radar axes are unscored because underlying data is insufficient; h-index (33) and i10-index (86) are quoted verbatim from the input snapshot without independent verification.
- Reproducibility (78) is averaged over only the 2 Evolve-reviewed papers and does not describe the full publication record.
- Recent catalog papers (LLM governance, efficient language models) diverge from the historical OpenAlex topic profile, and the 2026 annual count may reflect partial-year indexing.