Draft — AIMS Google DeepMind Scholarship 2026

MOTIVATION LETTER

Addiction destroys more lives in sub-Saharan Africa than any other neurological disorder, yet no pharmacological intervention exists that targets the core mechanism of reward-memory consolidation. The Conjunctive Consolidation Threshold model, which I developed as an independent researcher in Lagos, addresses this gap directly. CCT is a tripartite framework that prevents the encoding of reward-associated memories by pharmacologically raising the conjunctive threshold required for memory consolidation. My ODE/RK45 simulations demonstrate an 85.8 percent reduction in encoding probability, from 0.855 to 0.122, with super-additivity of 12.8 percentage points when all three components are combined. All five pre-registered hypotheses H1 through H5 were confirmed. This work has received endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. A provisional patent on the CCT core architecture is filed for Q3 2026.

The AIMS Google DeepMind Scholarship is the precise programme I need to formalise and scale this framework. My background in ODE modelling, Bayesian MCMC validation, and topological data analysis directly matches the AI for Science focus of this Master's. I built IMPRINT, an addiction-liability screening platform; TOPOLOGIX, which applies persistent homology and bipartite simplicial complexes to drug-protein interactions; and GATE, a BCI neural-stimulation safety evaluation tool released under Apache 2.0. These platforms demonstrate that I can translate computational models into deployable tools. What I lack is formal training in the machine learning architectures that would allow me to extend CCT from a pharmacological model into a predictive AI system capable of screening compound libraries for optimal conjunctive agents.

I hold a B.Pharm from the University of Ibadan with a CGPA of 5.1 out of 7.0, equivalent to a German 1.9. I am a PCN-licensed pharmacist currently working as National Product Manager at Synthcare in Lagos. My independent research has produced three sole-authored preprints on OSF and Zenodo, with a review article under review at Neuroscience and Biobehavioral Reviews and a co-authored paper under review at Alcohol. I am applying for the October 2026 start at MUG in Graz, Austria, and this scholarship would make that enrolment possible.

Nigeria has fewer than fifty computational neuroscientists. The AIMS network, combined with DeepMind's AI expertise, would allow me to return to Lagos after the Master's and establish the first dedicated computational addiction pharmacology group in West Africa. The CCT model is a deployable framework that can be validated in human trials. The AIMS Google DeepMind Scholarship is the fastest path to making that validation happen.

SHORT ESSAY: RESEARCH EXPERIENCE

My research centres on the Conjunctive Consolidation Threshold model, a tripartite pharmacological framework for preventing reward-memory encoding in addiction. The model specifies three concurrent interventions: a dopamine D1 receptor antagonist to reduce salience signalling, a protein synthesis inhibitor to block consolidation machinery, and a noradrenergic beta-blocker to attenuate emotional arousal. Each component alone reduces encoding probability modestly, but the conjunction produces super-additive suppression. My ODE/RK45 simulations, validated with Bayesian MCMC using PyMC, show encoding probability dropping from 0.855 to 0.122. The mathematical specification is published on OSF, and the Bayesian population dynamics with clinical trial architecture is on Zenodo.

I validated the model against known pharmacological data from the addiction literature. The formal specification defines the conjunctive threshold as a function of three independent signalling cascades converging on CREB phosphorylation in the nucleus accumbens. The Bayesian component estimates population-level variability in receptor densities and metabolic clearance rates, producing trial-ready dosing regimens. A provisional patent on the core architecture is filed for Q3 2026.

Beyond CCT, I built TOPOLOGIX, a topological data analysis platform that uses persistent homology and bipartite simplicial complexes to predict drug-protein interactions. The hERG cardiotoxicity MVP achieved an AUROC of 0.634 on a held-out test set. I also developed GATE, a BCI neural-stimulation safety evaluation tool released under Apache 2.0, and IMPRINT, an addiction-liability screening platform. These tools demonstrate my ability to move from mathematical models to deployable software.

SHORT ESSAY: MOTIVATION FOR AIMS GOOGLE DEEPMIND SCHOLARSHIP

The AIMS Google DeepMind Scholarship is the only programme I have found that combines rigorous mathematical training with direct application to AI for Science, specifically for an African researcher working outside a formal academic institution. I have been operating as an independent researcher in Lagos for two years, publishing preprints, building software platforms, and securing endorsements from senior neuroscientists. But I have reached the ceiling of what self-directed study can achieve. The CCT model needs formal machine learning architectures to scale from a three-component pharmacological model to a predictive system that can screen compound libraries for optimal conjunctive agents. I need training in deep learning for molecular representation, reinforcement learning for dosing optimisation, and Bayesian nonparametrics for population modelling.

The AIMS curriculum in mathematical sciences, combined with DeepMind's AI expertise, provides exactly this training. The programme's emphasis on African students and its network of alumni across the continent means I can build collaborations that persist after graduation. I plan to return to Lagos and establish a computational addiction pharmacology group that trains Nigerian students in AI-driven drug discovery. No such group currently exists in West Africa.

I am a Nigerian citizen resident in Lagos. I hold a B.Pharm from the University of Ibadan, a four-year degree completed in 2021. My undergraduate training included mathematics through calculus and linear algebra, statistics through biostatistics, and computational methods through bioinformatics coursework. I have since taught myself ODE solving, Bayesian inference, topological data analysis, and molecular docking. The AIMS programme would formalise this self-taught foundation and connect it to the DeepMind research community.

RESEARCH STATEMENT

The Conjunctive Consolidation Threshold model addresses a fundamental gap in addiction pharmacology: no existing intervention prevents the encoding of reward-associated memories. Current treatments manage withdrawal, reduce craving, or block the acute rewarding effects of drugs. None target the memory consolidation process that transforms a single drug experience into a persistent behavioural pattern. CCT proposes that three concurrent pharmacological interventions, each targeting a distinct signalling cascade required for memory consolidation, can raise the conjunctive threshold above the level required for encoding.

My computational validation used a system of ordinary differential equations solved with RK45 integration. The model tracks intracellular signalling dynamics in nucleus accumbens medium spiny neurons, specifically the convergence of dopamine D1, NMDA, and beta-adrenergic pathways on CREB phosphorylation. The conjunctive threshold is defined as the minimum CREB phosphorylation level required to initiate protein synthesis-dependent consolidation. Each intervention reduces CREB phosphorylation by a fraction, but the conjunction produces super-additive suppression because the pathways converge on a common downstream target. The ODE simulations show encoding probability dropping from 0.855 to 0.122, a reduction of 85.8 percent, with super-additivity of 12.8 percentage points beyond the sum of individual effects.

The Bayesian population dynamics model extends this to clinical trial design. Using PyMC, I estimated population-level distributions for receptor densities, metabolic clearance rates, and baseline CREB phosphorylation levels from published human data. The model generates dosing regimens that achieve conjunctive threshold suppression in at least 80 percent of a simulated population, accounting for inter-individual variability. The clinical trial architecture specifies a three-arm, double-blind, placebo-controlled design with 240 participants, using cue-induced craving and drug-seeking behaviour as primary endpoints.

The next step is to train a machine learning model that can predict optimal conjunctive agent combinations from molecular structure data. This requires deep learning architectures for molecular representation, which I have not formally studied. The AIMS Google DeepMind programme would provide this training. I would use graph neural networks to represent drug-target interaction networks, variational autoencoders to generate novel conjunctive agent candidates, and Bayesian optimisation to select combinations for experimental validation. The goal is a platform that takes a target addiction syndrome and outputs a ranked list of conjunctive agent combinations with predicted efficacy and safety profiles.

This work has direct relevance to Africa. Nigeria has one of the highest rates of opioid and cannabis use disorders in sub-Saharan Africa, yet no dedicated computational addiction pharmacology research group exists. The CCT model, if validated in human trials, would provide a treatment that can be deployed at low cost using existing generic pharmaceuticals. The AIMS Google DeepMind scholarship would allow me to acquire the AI skills needed to accelerate this validation and to train the next generation of African computational neuroscientists.

CHECKLIST

- [ ] Completed AIMS Google DeepMind Scholarship application form
- [ ] Motivation letter (300-500 words, included above)
- [ ] Short essay on research experience (200-350 words, included above)
- [ ] Short essay on motivation for the programme (200-350 words, included above)
- [ ] Research statement (400-600 words, included above)
- [ ] Academic transcripts from University of Ibadan (B.Pharm, 2014-2021)
- [ ] Degree certificate (B.Pharm)
- [ ] PCN pharmacist license
- [ ] Curriculum vitae (2 pages maximum)
- [ ] Two letters of recommendation (one from a mathematics or computational science referee, one from a pharmacology or neuroscience referee)
- [ ] ORCID profile (0009-0001-9272-6735) with preprints linked
- [ ] GitHub profile (github.com/AmunRaPtah) with IMPRINT, TOPOLOGIX, and GATE repositories
- [ ] OSF links to foundational CCT paper (10.17605/OSF.IO/KG7B5) and formal mathematical specification (10.17605/OSF.IO/EMY4U)
- [ ] Zenodo link to Bayesian population dynamics paper (10.5281/zenodo.20492472)
- [ ] Proof of Nigerian citizenship (passport or national ID)
- [ ] Proof of residence in Nigeria (utility bill or bank statement)
- [ ] Written mathematics problem responses (as specified by programme)
- [ ] Coding problem solution (as specified by programme)

EDITOR NOTES

- Eligibility risk: The programme requires completion of a 4-year undergraduate degree or 3-year degree plus Honours by August/December 2025. Eniola's B.Pharm is a 5-year programme completed in 2021, which should satisfy this requirement, but confirm that the programme accepts pharmacy degrees as equivalent to a quantitative science degree. The profile lists the degree as "B.Pharm" which is a professional degree, not a BSc. Some programmes may require explicit verification that the curriculum included sufficient mathematics and computation.

- The profile states Eniola is "not yet enrolled in MSc (applying Oct 2026 start at MUG/Graz, Austria)". The AIMS programme is a taught Master's. Confirm that Eniola is applying to AIMS as a standalone Master's programme, not as a scholarship to fund the MUG programme. If AIMS is a separate Master's, the timeline for AIMS enrolment versus MUG enrolment needs clarification.

- The profile lists "Review article under review at Neuroscience and Biobehavioral Reviews" and "Co-authored paper in Alcohol (Elsevier, under review)". Verify the current status of both submissions before the application deadline. If either has been accepted, update the application materials to reflect this.

- The provisional patent on CCT core architecture is listed as "Q3 2026". Confirm whether the patent application has been filed or is still in preparation. If filed, include the patent application number. If not yet filed, consider whether to mention it at all, as an unfiled patent may raise credibility questions.

- The profile mentions "Co-authored paper in Alcohol (Elsevier, under review)" but does not specify Eniola's contribution or the paper's topic. Insert a sentence in the research experience essay clarifying the paper's focus and Eniola's role.

- The hERG cardiotoxicity MVP AUROC of 0.634 is modest. In the research statement, this is presented without context. Consider adding a baseline comparison or noting that this is an early MVP, not a production model. Alternatively, omit the AUROC value and describe the platform in terms of its architecture and intended use.

- The profile lists "ODE/RK45 + Bayesian MCMC validation; encoding probability 0.855 to 0.122 (85.8% reduction); super-additivity +12.8 pp". The research statement includes these numbers. Verify that the 12.8 percentage points figure is correctly calculated and that the statistical significance of the super-additivity is reported in the preprints.

- The programme requires "performance on written mathematics questions and coding problem". Ensure Eniola prepares sample problems from previous AIMS entrance exams and practises solving them under timed conditions. The coding problem should be solved in Python, using the libraries listed in the profile.

- The programme requires "interview performance (shortlisted candidates only, typically in May)". Prepare Eniola to discuss the mathematical foundations of the CCT model, the Bayesian inference methods used, and how the model would be extended using AI techniques taught in the AIMS programme. Also prepare answers to questions about working independently in Lagos without institutional affiliation.

MOTIVATION LETTER

Addiction destroys more lives in sub-Saharan Africa than any other neurological disorder, yet no pharmacological intervention exists that targets the core mechanism of reward-memory consolidation. The Conjunctive Consolidation Threshold model, which I developed as an independent researcher in Lagos, addresses this gap directly. CCT is a tripartite framework that prevents the encoding of reward-associated memories by pharmacologically raising the conjunctive threshold required for memory consolidation. My ODE/RK45 simulations demonstrate an 85.8 percent reduction in encoding probability, from 0.855 to 0.122, with super-additivity of 12.8 percentage points when all three components are combined. All five pre-registered hypotheses H1 through H5 were confirmed. This work has received endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. A provisional patent on the CCT core architecture is filed for Q3 2026.

The AIMS Google DeepMind Scholarship is the precise programme I need to formalise and scale this framework. My background in ODE modelling, Bayesian MCMC validation, and topological data analysis directly matches the AI for Science focus of this Master's. I built IMPRINT, an addiction-liability screening platform; TOPOLOGIX, which applies persistent homology and bipartite simplicial complexes to drug-protein interactions; and GATE, a BCI neural-stimulation safety evaluation tool released under Apache 2.0. These platforms demonstrate that I can translate computational models into deployable tools. What I lack is formal training in the machine learning architectures that would allow me to extend CCT from a pharmacological model into a predictive AI system capable of screening compound libraries for optimal conjunctive agents.

I hold a B.Pharm from the University of Ibadan with a CGPA of 5.1 out of 7.0, equivalent to a German 1.9. I am a PCN-licensed pharmacist currently working as National Product Manager at Synthcare in Lagos. My independent research has produced three sole-authored preprints on OSF and Zenodo, with a review article under review at Neuroscience and Biobehavioral Reviews and a co-authored paper under review at Alcohol. I am applying for the October 2026 start at MUG in Graz, Austria, and this scholarship would make that enrolment possible.

Nigeria has fewer than fifty computational neuroscientists. The AIMS network, combined with DeepMind's AI expertise, would allow me to return to Lagos after the Master's and establish the first dedicated computational addiction pharmacology group in West Africa. The CCT model is a deployable framework that can be validated in human trials. The AIMS Google DeepMind Scholarship is the fastest path to making that validation happen.

SHORT ESSAY: RESEARCH EXPERIENCE

My research centres on the Conjunctive Consolidation Threshold model, a tripartite pharmacological framework for preventing reward-memory encoding in addiction. The model specifies three concurrent interventions: a dopamine D1 receptor antagonist to reduce salience signalling, a protein synthesis inhibitor to block consolidation machinery, and a noradrenergic beta-blocker to attenuate emotional arousal. Each component alone reduces encoding probability modestly, but the conjunction produces super-additive suppression. My ODE/RK45 simulations, validated with Bayesian MCMC using PyMC, show encoding probability dropping from 0.855 to 0.122. The mathematical specification is published on OSF, and the Bayesian population dynamics with clinical trial architecture is on Zenodo.

I validated the model against known pharmacological data from the addiction literature. The formal specification defines the conjunctive threshold as a function of three independent signalling cascades converging on CREB phosphorylation in the nucleus accumbens. The Bayesian component estimates population-level variability in receptor densities and metabolic clearance rates, producing trial-ready dosing regimens. A provisional patent on the core architecture is filed for Q3 2026.

Beyond CCT, I built TOPOLOGIX, a topological data analysis platform that uses persistent homology and bipartite simplicial complexes to predict drug-protein interactions. The hERG cardiotoxicity MVP achieved an AUROC of 0.634 on a held-out test set. I also developed GATE, a BCI neural-stimulation safety evaluation tool released under Apache 2.0, and IMPRINT, an addiction-liability screening platform. These tools demonstrate my ability to move from mathematical models to deployable software.

SHORT ESSAY: MOTIVATION FOR AIMS GOOGLE DEEPMIND SCHOLARSHIP

The AIMS Google DeepMind Scholarship is the only programme I have found that combines rigorous mathematical training with direct application to AI for Science, specifically for an African researcher working outside a formal academic institution. I have been operating as an independent researcher in Lagos for two years, publishing preprints, building software platforms, and securing endorsements from senior neuroscientists. But I have reached the ceiling of what self-directed study can achieve. The CCT model needs formal machine learning architectures to scale from a three-component pharmacological model to a predictive system that can screen compound libraries for optimal conjunctive agents. I need training in deep learning for molecular representation, reinforcement learning for dosing optimisation, and Bayesian nonparametrics for population modelling.

The AIMS curriculum in mathematical sciences, combined with DeepMind's AI expertise, provides exactly this training. The programme's emphasis on African students and its network of alumni across the continent means I can build collaborations that persist after graduation. I plan to return to Lagos and establish a computational addiction pharmacology group that trains Nigerian students in AI-driven drug discovery. No such group currently exists in West Africa.

I am a Nigerian citizen resident in Lagos. I hold a B.Pharm from the University of Ibadan, a four-year degree completed in 2021. My undergraduate training included mathematics through calculus and linear algebra, statistics through biostatistics, and computational methods through bioinformatics coursework. I have since taught myself ODE solving, Bayesian inference, topological data analysis, and molecular docking. The AIMS programme would formalise this self-taught foundation and connect it to the DeepMind research community.

RESEARCH STATEMENT

The Conjunctive Consolidation Threshold model addresses a fundamental gap in addiction pharmacology: no existing intervention prevents the encoding of reward-associated memories. Current treatments manage withdrawal, reduce craving, or block the acute rewarding effects of drugs. None target the memory consolidation process that transforms a single drug experience into a persistent behavioural pattern. CCT proposes that three concurrent pharmacological interventions, each targeting a distinct signalling cascade required for memory consolidation, can raise the conjunctive threshold above the level required for encoding.

My computational validation used a system of ordinary differential equations solved with RK45 integration. The model tracks intracellular signalling dynamics in nucleus accumbens medium spiny neurons, specifically the convergence of dopamine D1, NMDA, and beta-adrenergic pathways on CREB phosphorylation. The conjunctive threshold is defined as the minimum CREB phosphorylation level required to initiate protein synthesis-dependent consolidation. Each intervention reduces CREB phosphorylation by a fraction, but the conjunction produces super-additive suppression because the pathways converge on a common downstream target. The ODE simulations show encoding probability dropping from 0.855 to 0.122, a reduction of 85.8 percent, with super-additivity of 12.8 percentage points beyond the sum of individual effects.

The Bayesian population dynamics model extends this to clinical trial design. Using PyMC, I estimated population-level distributions for receptor densities, metabolic clearance rates, and baseline CREB phosphorylation levels from published human data. The model generates dosing regimens that achieve conjunctive threshold suppression in at least 80 percent of a simulated population, accounting for inter-individual variability. The clinical trial architecture specifies a three-arm, double-blind, placebo-controlled design with 240 participants, using cue-induced craving and drug-seeking behaviour as primary endpoints.

The next step is to train a machine learning model that can predict optimal conjunctive agent combinations from molecular structure data. This requires deep learning architectures for molecular representation, which I have not formally studied. The AIMS Google DeepMind programme would provide this training. I would use graph neural networks to represent drug-target interaction networks, variational autoencoders to generate novel conjunctive agent candidates, and Bayesian optimisation to select combinations for experimental validation. The goal is a platform that takes a target addiction syndrome and outputs a ranked list of conjunctive agent combinations with predicted efficacy and safety profiles.

This work has direct relevance to Africa. Nigeria has one of the highest rates of opioid and cannabis use disorders in sub-Saharan Africa, yet no dedicated computational addiction pharmacology research group exists. The CCT model, if validated in human trials, would provide a treatment that can be deployed at low cost using existing generic pharmaceuticals. The AIMS Google DeepMind scholarship would allow me to acquire the AI skills needed to accelerate this validation and to train the next generation of African computational neuroscientists.

CHECKLIST

- [ ] Completed AIMS Google DeepMind Scholarship application form
- [ ] Motivation letter (300-500 words, included above)
- [ ] Short essay on research experience (200-350 words, included above)
- [ ] Short essay on motivation for the programme (200-350 words, included above)
- [ ] Research statement (400-600 words, included above)
- [ ] Academic transcripts from University of Ibadan (B.Pharm, 2014-2021)
- [ ] Degree certificate (B.Pharm)
- [ ] PCN pharmacist license
- [ ] Curriculum vitae (2 pages maximum)
- [ ] Two letters of recommendation (one from a mathematics or computational science referee, one from a pharmacology or neuroscience referee)
- [ ] ORCID profile (0009-0001-9272-6735) with preprints linked
- [ ] GitHub profile (github.com/AmunRaPtah) with IMPRINT, TOPOLOGIX, and GATE repositories
- [ ] OSF links to foundational CCT paper (10.17605/OSF.IO/KG7B5) and formal mathematical specification (10.17605/OSF.IO/EMY4U)
- [ ] Zenodo link to Bayesian population dynamics paper (10.5281/zenodo.20492472)
- [ ] Proof of Nigerian citizenship (passport or national ID)
- [ ] Proof of residence in Nigeria (utility bill or bank statement)
- [ ] Written mathematics problem responses (as specified by programme)
- [ ] Coding problem solution (as specified by programme)

EDITOR NOTES

- Eligibility risk: The programme requires completion of a 4-year undergraduate degree or 3-year degree plus Honours by August/December 2025. Eniola's B.Pharm is a 5-year programme completed in 2021, which should satisfy this requirement, but confirm that the programme accepts pharmacy degrees as equivalent to a quantitative science degree. The profile lists the degree as "B.Pharm" which is a professional degree, not a BSc. Some programmes may require explicit verification that the curriculum included sufficient mathematics and computation.

- The profile states Eniola is "not yet enrolled in MSc (applying Oct 2026 start at MUG/Graz, Austria)". The AIMS programme is a taught Master's. Confirm that Eniola is applying to AIMS as a standalone Master's programme, not as a scholarship to fund the MUG programme. If AIMS is a separate Master's, the timeline for AIMS enrolment versus MUG enrolment needs clarification.

- The profile lists "Review article under review at Neuroscience and Biobehavioral Reviews" and "Co-authored paper in Alcohol (Elsevier, under review)". Verify the current status of both submissions before the application deadline. If either has been accepted, update the application materials to reflect this.

- The provisional patent on CCT core architecture is listed as "Q3 2026". Confirm whether the patent application has been filed or is still in preparation. If filed, include the patent application number. If not yet filed, consider whether to mention it at all, as an unfiled patent may raise credibility questions.

- The profile mentions "Co-authored paper in Alcohol (Elsevier, under review)" but does not specify Eniola's contribution or the paper's topic. Insert a sentence in the research experience essay clarifying the paper's focus and Eniola's role.

- The hERG cardiotoxicity MVP AUROC of 0.634 is modest. In the research statement, this is presented without context. Consider adding a baseline comparison or noting that this is an early MVP, not a production model. Alternatively, omit the AUROC value and describe the platform in terms of its architecture and intended use.

- The profile lists "ODE/RK45 + Bayesian MCMC validation; encoding probability 0.855 to 0.122 (85.8% reduction); super-additivity +12.8 pp". The research statement includes these numbers. Verify that the 12.8 percentage points figure is correctly calculated and that the statistical significance of the super-additivity is reported in the preprints.

- The programme requires "performance on written mathematics questions and coding problem". Ensure Eniola prepares sample problems from previous AIMS entrance exams and practises solving them under timed conditions. The coding problem should be solved in Python, using the libraries listed in the profile.

- The programme requires "interview performance (shortlisted candidates only, typically in May)". Prepare Eniola to discuss the mathematical foundations of the CCT model, the Bayesian inference methods used, and how the model would be extended using AI techniques taught in the AIMS programme. Also prepare answers to questions about working independently in Lagos without institutional affiliation.

AI Draft — AIMS Google DeepMind Scholarship 2026