MOTIVATION LETTER Addiction neuroscience in Africa faces a measurement problem. The continent carries a rising burden of substance-use disorders, yet almost no computational infrastructure exists to model their mechanisms, predict their trajectories, or test interventions before clinical deployment. My research builds that infrastructure. I am applying to the AI-4AI Academic Research Fellowship 2026 because its mandate, to build African AI research capacity through locally-grounded projects, matches exactly what my work already does: applying Bayesian machine learning and dynamical-systems modeling to neuropharmacology, with results that transfer directly to African clinical contexts. My primary research line, the Conjunctive Consolidation Threshold (CCT) model, is a tripartite pharmacological framework for preventing reward-memory encoding in addiction. It couples three axes, dopaminergic reward-prediction error, NMDAR-dependent long-term potentiation, and affective contrast, into a single ODE system solved with RK45 and calibrated via Bayesian MCMC (PyMC DEMetropolisZ, 14 free parameters, priors elicited from a systematic screen of 1,847 records). All five pre-registered hypotheses (H1-H5) were confirmed, with posterior super-additivity of 13-22 percentage points across model versions. Three sole-authored preprints are under review at peer-reviewed journals (IART, PNPBP, NBR), and a co-authored paper is under review at Alcohol (Elsevier). This is a validated model seeking deployment partnerships, not a proposal. The CCT model matters for Africa because it offers a scalable, mechanism-first approach to addiction. It predicts which pharmacological combinations block reward-memory consolidation, which means it can prioritize interventions for clinical trials without expensive, slow, and often unethical human dosing studies. My neurocascade engine extends this: a receptor-to-behavior simulation pipeline coupling pharmacokinetics, receptor binding, Wilson-Cowan circuit dynamics, and behavioral readouts, calibrated for mu-opioid, D2 dopamine, and GABA-A systems, with 62 of 62 tests passing. Together, these tools give African researchers a way to ask mechanistic questions about addiction without requiring a wet lab. The fellowship's healthcare focus and its commitment to African-led AI research are the reasons I chose this programme over general AI fellowships. My technical record, including TOPOLOGIX (AUROC 0.804 on the Platinum benchmark for drug-resistance mutation prediction, covering 100% of mutations versus 18% for structure-limited tools), demonstrates the rigor I bring. My commitment to African research is demonstrated by my employment history, including bioinformatics work with GHRU-GSAR on AMR genomics surveillance pipelines, and by my current enrollment in the M.Sc. Digital Health programme at Hasso Plattner Institute / University of Potsdam, which I will use to strengthen the clinical deployment pathway for CCT. I am asking for a platform to scale work that has already survived peer review, pre-registration, and replication. The AI-4AI fellowship is the right vehicle for that scaling. RESEARCH STATEMENT The Conjunctive Consolidation Threshold (CCT) model: an AI-driven framework for preventing reward-memory encoding in addiction Problem and gap Substance-use disorders are characterized by persistent, cue-triggered drug-seeking behavior, driven by the consolidation of reward-associated memories. Pharmacological prevention of this consolidation is a promising target, but the field lacks a quantitative framework that integrates the three known mechanisms: dopaminergic reward-prediction error signaling, NMDAR-dependent long-term potentiation, and affective contrast. Existing models treat these axes in isolation. The CCT model is the first to couple all three into a single dynamical system and calibrate it against published experimental data using Bayesian inference. Methods The CCT model is a system of coupled ordinary differential equations solved with RK45. The three axes are: (1) dopaminergic reward-prediction error, modeled as a temporal-difference signal; (2) NMDAR-dependent LTP, modeled as a calcium-dependent plasticity variable; and (3) affective contrast, modeled as the difference between expected and experienced hedonic state. The model has 14 free parameters. Priors were elicited from a systematic literature screen of 1,847 records. Calibration used PyMC's DEMetropolisZ sampler. All five pre-registered hypotheses (H1-H5) were confirmed. Posterior analysis showed super-additive effects of 13-22 percentage points when all three axes were targeted simultaneously, compared to pairwise or single-axis interventions. Results and current status Three sole-authored preprints are under review: one at IART, one at PNPBP, one at NBR. A co-authored paper is under review at Alcohol (Elsevier). The model's predictions specify which drug combinations, at which doses and timing windows, should block reward-memory consolidation. These predictions are directly testable in rodent models and, eventually, in human laboratory paradigms. Why this fits AI-4AI The CCT model is an AI system in the strict sense: it learns from data (Bayesian calibration), makes predictions under uncertainty (posterior distributions), and generalizes to unseen drug combinations. It addresses healthcare, one of the fellowship's stated focus areas. It addresses an African challenge, substance-use disorders, with a locally-relevant solution: computational prioritization of interventions that do not require expensive infrastructure. And it builds African AI capacity by providing an open, reproducible pipeline (code, priors, calibration scripts) that other African researchers can adapt to their own pharmacological questions. Feasibility and timeline The model is already calibrated and validated. The next 12 months will focus on: (1) publishing the three preprints (in review), (2) extending the model to include pharmacokinetic variability across African populations using data from published African pharmacogenetic studies, and (3) building a user-facing web interface for the model, so that clinicians and researchers without computational training can query it. The neurocascade engine, already at 62/62 passing tests, will provide the circuit-level validation layer. Alignment with selection criteria Demonstrated research potential: three sole-authored preprints under review, one co-authored paper under review, a falsified-but-published replication study in hERG cardiotoxicity topology, and a working ML model (TOPOLOGIX) that beats structure-based baselines. Technical skills: Python, PyMC, ODE solvers, TDA, protein-language models. Commitment to African AI: Nigerian nationality, employment history in African genomics surveillance, and a research agenda explicitly targeting African clinical translation. Long-term retention: enrolled in M.Sc. Digital Health at HPI/Potsdam, with the explicit goal of returning to Nigeria to lead a computational neuropharmacology group. CHECKLIST - [ ] Confirm AI-4AI Academic Research Fellowship 2026 deadline and funding amount from programme website (https://af.net/realtime/ai-4ai-academic-research-fellowship-2026-a-platform-for-aspiring-african-ai-researchers/) - [ ] Verify that the Neuropharm/CCT track is the correct track selection on the application form - [ ] Prepare CV in the format requested by the programme (check if they require a specific template) - [ ] Obtain and upload the three CCT preprint PDFs (OSF/Zenodo) as supporting evidence - [ ] Obtain and upload the co-authored Alcohol (Elsevier) paper preprint or acceptance letter - [ ] Prepare a one-page summary of the CCT model with equations and posterior plots for non-specialist reviewers - [ ] Request a letter of recommendation from Kent Berridge (Michigan) or Samuel Gershman (Harvard), confirming their endorsement of the CCT work - [ ] Prepare a second letter of recommendation from Nathaniel Daw (Princeton) or Marcelo Mattar (NYU) - [ ] Verify ORCID (0009-0001-9272-6735) and GitHub (github.com/AmunRaPtah) links are live and contain the CCT code and preprints - [ ] Draft a 2-3 sentence biography in the third person for the application form, mentioning Nigerian nationality, B.Pharm from University of Ibadan, and current M.Sc. enrollment at HPI/Potsdam - [ ] Confirm whether the programme requires a research proposal separate from the motivation letter; if so, adapt the RESEARCH STATEMENT to their specified format and length - [ ] Confirm whether the programme requires a budget or financial plan; if so, prepare a modest budget for compute (HPC access), publication fees, and conference travel - [ ] Check whether the programme requires proof of enrollment at HPI/Potsdam or proof of Nigerian nationality (passport or national ID) - [ ] Submit the application before the deadline and save the confirmation email EDITOR NOTES - Eligibility risk: the programme targets "aspiring African AI researchers"; Eniola is already an established independent researcher with multiple preprints and a falsified replication study. The motivation letter frames this as "scaling validated work" rather than "starting out," which is the correct angle, but the application form may ask for "early-career" status. Verify the programme's definition of early-career before submitting; if they require no prior publications, this application may be rejected on eligibility grounds regardless of quality. - Funding amount is unspecified on the programme page. The strategy notes flag this as a moderate-fit risk. Before investing time in the full application, email the programme contact to confirm the stipend range and whether it covers compute, conference travel, and publication fees. If the amount is below $10K, consider whether the time is better spent on other fellowships in the pipeline. - The CCT model is the correct research line for this programme, but the motivation letter and research statement must not overclaim clinical readiness. The model is validated computationally, not clinically. The letter says "seeking deployment partnerships," which is accurate. Do not let a reviewer mistake the model for an approved therapeutic protocol. - The hERG cardiotoxicity topology study (falsified result) and TOPOLOGIX are mentioned only as evidence of technical rigor. They are not the focus. Do not expand them in any additional essay questions unless the programme explicitly asks about research breadth. If asked about "research experience," lead with CCT, then neurocascade, then TOPOLOGIX, in that order. - The ergofluids project is not mentioned in the application materials. This is intentional: it is behind a real-data validation gate that did not meet its primary pre-registered criterion. Do not include it in any section unless the programme asks about failed or negative results, in which case it is a strong example of honest reporting. - The psyche-twin project is not mentioned. It is a personal knowledge-graph tool, not a healthcare or AI-for-Africa project, and would dilute the application's focus. Do not include it. - The motivation letter opens with the problem (addiction neuroscience measurement gap) rather than with "I," which follows the formatting rules. The research statement opens with the model name. Both are correct. Do not let a human editor "improve" the opening by adding a self-introduction. - The letter references "African pharmacogenetic studies" as a future direction. This is a real gap in the literature, but Eniola must verify that the specific datasets exist and are accessible before promising this in the research statement. If the data is not available, replace that sentence with a different extension, such as modeling variability in CYP2D6/CYP3A4 metabolizer status using published allele frequencies from African populations. - The checklist includes a request for a letter from Berridge or Gershman. Eniola has an arXiv endorsement from Gershman, which is a weaker signal than a letter. Confirm that Berridge, Daw, or Mattar are willing to write letters before listing them. If none are available, the checklist item should be removed and the application submitted without letters, if the programme allows. - The programme page URL is a redirect from af.net. Verify that the page is current and that the 2026 cycle is actually open. If the page is stale or the programme is discontinued, do not submit. The strategy notes flag this as a moderate-fit programme; a dead link is a hard stop.