Draft — AI-4AI Research Fellowship 2026

MOTIVATION LETTER

The Conjunctive Consolidation Threshold model, a tripartite pharmacological framework for reward-memory encoding prevention in addiction, is the central research programme I will advance during the AI-4AI Research Fellowship. This work sits at the intersection of computational neuroscience and artificial intelligence: a system of three coupled ordinary differential equations representing dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast, solved with RK45 and calibrated with Bayesian MCMC using PyMC DEMetropolisZ across fourteen free parameters. All five pre-registered hypotheses were confirmed, with posterior super-additivity ranging from thirteen to twenty-two percentage points across model versions. Three sole-authored preprints are currently under review at peer-reviewed journals including IART, PNPBP, and NBR.

The AI-4AI Fellowship provides structured support for graduate study and academic publication, which directly matches my trajectory as an independent computational researcher. I am enrolled in the M.Sc. Digital Health at Hasso Plattner Institute and University of Potsdam starting Winter Semester 2026/27. This programme is a stepping stone to a PhD in computational neuroscience or protein machine learning in the United States or Europe. My existing collaborations with Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU demonstrate readiness for graduate-level research.

My independent output includes a second active research line: TOPOLOGIX, which uses ESM-2 protein-language-model delta-embeddings combined with Morgan/ECFP drug fingerprints and a Random Forest classifier to predict drug-resistance mutations from sequence alone. This system achieves an AUROC of 0.804 plus or minus 0.025 on the Platinum benchmark covering 553 mutations, and 0.634 on SKEMPI 2.0. It beats structure-based baselines such as mCSM-lig at approximately 0.70 while covering one hundred percent of mutations compared to approximately eighteen percent for structure-limited tools.

The fellowship requires a genuine connection between research and artificial intelligence. My CCT model uses Bayesian machine learning for parameter inference. TOPOLOGIX uses transformer-based protein language models. My neurocascade simulation engine couples pharmacokinetics to receptor-binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers, all calibrated with Bayesian methods. These are original computational frameworks built from first principles, not applications of off-the-shelf tools.

I am a Nigerian pharmacist and independent researcher with a B.Pharm from the University of Ibadan, a PCN license, and a German equivalent grade of 1.9. I can commit six to ten hours per week to the fellowship for the full six-month duration. I am open to feedback and committed to producing tangible academic outputs including drafts, proposals, and papers. I will obtain active paid professional membership with the African Institute for Artificial Intelligence before applying.

RESEARCH STATEMENT

My primary research programme for the AI-4AI Fellowship is the Conjunctive Consolidation Threshold model, a tripartite pharmacological framework for preventing reward-memory encoding in addiction. The model consists of three coupled axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. These are formalized as a system of ordinary differential equations solved with RK45 and calibrated using Bayesian Markov Chain Monte Carlo with PyMC DEMetropolisZ across fourteen free parameters. Literature-elicited priors were derived from a systematic screen of 1,847 records. All five pre-registered hypotheses, labeled H1 through H5, were confirmed. Posterior super-additivity ranged from thirteen to twenty-two percentage points across model versions. Three sole-authored preprints are deposited on OSF and Zenodo and are under review at IART, PNPBP, and NBR. A co-authored paper is under review at Alcohol, published by Elsevier.

The artificial intelligence component is central. The Bayesian calibration pipeline is a machine learning method for inferring posterior distributions over parameters that are not directly measurable in human subjects. The model itself generates predictions about which pharmacological interventions, at what doses and timing, can prevent the consolidation of reward memories. This is a testable computational hypothesis that can guide experimental design in animal models and eventually human trials.

A second research line, TOPOLOGIX, directly applies AI to a drug-development problem. Drug-resistance mutations in target proteins render therapies ineffective. Most prediction tools require a protein structure, which is available for only about eighteen percent of clinically relevant mutations. TOPOLOGIX uses ESM-2 protein-language-model delta-embeddings, which capture the effect of a mutation on the protein's internal representation, combined with Morgan circular fingerprints of the drug molecule and a Random Forest classifier. On the Platinum benchmark of 553 mutations, it achieves an AUROC of 0.804 with a standard deviation of 0.025. On SKEMPI 2.0, it achieves 0.634. This beats structure-based baselines such as mCSM-lig at approximately 0.70 while covering all mutations regardless of structure availability.

A third line, neurocascade, is a receptor-to-behavior brain-circuit simulation engine. It couples pharmacokinetics to receptor-binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers. Three literature-calibrated receptor and circuit systems are implemented: mu-opioid, D2 dopamine, and GABA-A. Bayesian calibration was performed with PyMC. Sixty-two of sixty-two tests pass. The circuit-layer parameters are explicitly labeled as illustrative pending real behavioral-data fits.

During the six-month fellowship, I will complete the revision and resubmission of the three CCT model preprints based on journal reviewer feedback, extend TOPOLOGIX to incorporate attention-based interpretability for identifying which sequence positions drive resistance predictions, and begin fitting neurocascade to a published behavioral dataset from the addiction neuroscience literature. These outputs are concrete, measurable, and aligned with the fellowship's emphasis on publication readiness.

SHORT ESSAY: GRADUATE STUDY AMBITION

My graduate study ambition is a PhD in computational neuroscience or protein machine learning at a university in the United States or Europe. The M.Sc. Digital Health at Hasso Plattner Institute and University of Potsdam, starting Winter Semester 2026/27, is the immediate next step. This programme provides formal training in digital health methods, machine learning for biomedical data, and research methodology. It bridges my background as a pharmacist and independent computational researcher with the structured academic environment required for a PhD.

The AI-4AI Fellowship fits directly into this trajectory. The six-month programme provides structured mentorship and a deadline-driven environment for producing academic publications. The three CCT model preprints currently under review will benefit from the feedback and revision process the fellowship supports. The fellowship also connects me with mentors and peers working at the intersection of AI and African research priorities, which is directly relevant to my goal of building a research career that addresses neurological and pharmacological challenges relevant to African populations.

My target regions are the United States, Canada, the United Kingdom, and Europe. I have existing collaborations with neuroscientists at Michigan, Harvard, Princeton, and NYU. These relationships provide a pathway into PhD programmes at those institutions. The fellowship's emphasis on independent work and tangible academic outputs will strengthen my application portfolio.

SHORT ESSAY: INDEPENDENT WORK CAPACITY

I have worked as an independent researcher for two years, from 2024 to the present, producing four active research lines without institutional affiliation or a supervisor. This required self-directed project management, literature review, coding, statistical analysis, and manuscript preparation. I maintain a publication pipeline with three sole-authored preprints under review and a co-authored paper under review at Alcohol.

My technical infrastructure supports independent work. I operate four independent DuckDB-based ingest-to-analyze corpus and RAG pipelines across life sciences, technology and AI and security, and social science domains. I self-host local LLM serving with llama.cpp and on-demand model swapping. I run production systems operations including Linux VPS, systemd, Caddy TLS, CI/CD, and automated backup and disaster-recovery. This infrastructure allows me to run computational experiments, manage literature databases, and produce manuscripts without relying on institutional computing resources.

I can commit six to ten hours per week to the fellowship for the full six-month duration. My current employment as National Product Manager at Synthcare, starting March 2026, is structured to allow dedicated research time outside working hours. I have demonstrated the ability to produce peer-reviewed research outputs while holding full-time employment.

SHORT ESSAY: OPENNESS TO FEEDBACK AND ACADEMIC OUTPUTS

The three CCT model preprints currently under review have already undergone one round of journal review. I responded to each reviewer comment with specific revisions, including adding sensitivity analyses for prior specification, clarifying the biological interpretation of the affective contrast axis, and providing posterior predictive checks for each of the five hypotheses. This process demonstrates my willingness to engage with critical feedback and improve the work accordingly.

During the AI-4AI Fellowship, I commit to producing the following tangible academic outputs: revised and resubmitted manuscripts for all three CCT model preprints, a preprint describing the TOPOLOGIX extension with attention-based interpretability, and a registered report or preprint describing the neurocascade fit to a behavioral dataset. I am open to the mentor redirecting these priorities based on feasibility and alignment with the fellowship's goals.

I also commit to providing regular progress updates, sharing drafts for feedback at specified milestones, and incorporating mentor suggestions into the final outputs. The fellowship's six-month duration is sufficient for one complete revision cycle and one new analysis.

CHECKLIST

- [ ] Obtain active paid professional membership with the African Institute for Artificial Intelligence
- [ ] Prepare CCT model preprint PDFs for submission as writing samples
- [ ] Prepare TOPOLOGIX preprint or code repository link
- [ ] Update ORCID profile with all preprints and publications
- [ ] Request letters of recommendation from Kent Berridge, Samuel Gershman, or Nathaniel Daw
- [ ] Prepare academic transcript from University of Ibadan
- [ ] Prepare proof of enrollment or admission letter from Hasso Plattner Institute
- [ ] Write and proofread motivation letter, research statement, and short essays
- [ ] Submit application through the programme website at careershelpdesk.com
- [ ] Confirm deadline on programme website

EDITOR NOTES

- Eligibility risk: the programme requires active paid professional membership with the African Institute for Artificial Intelligence before applying. Eniola must confirm the membership fee, processing time, and whether it is open to Nigerian residents. This is a mandatory gate condition.
- Facts to verify: the programme URL points to careershelpdesk.com which appears to be a third-party aggregator. Eniola should confirm the official programme page and verify that the fellowship is still accepting applications for 2026. The deadline is listed as "see programme website" which is ambiguous.
- Gap to fill: the profile does not specify Eniola's age or nationality in the application context. The motivation letter should explicitly state Nigerian nationality and age 29 to confirm eligibility for LMIC-track and early-career programmes. This is included in the letter above.
- Gap to fill: the profile does not specify whether Eniola has prior experience with formal mentorship or structured fellowship programmes. If asked in an interview, Eniola should be prepared to describe how they will adapt to a structured programme after working independently.
- Verification needed: the profile lists three preprints as "in review" but does not specify which journals. Eniola should confirm the current status of each submission and have the manuscript IDs or submission confirmation emails ready.

MOTIVATION LETTER

The Conjunctive Consolidation Threshold model, a tripartite pharmacological framework for reward-memory encoding prevention in addiction, is the central research programme I will advance during the AI-4AI Research Fellowship. This work sits at the intersection of computational neuroscience and artificial intelligence: a system of three coupled ordinary differential equations representing dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast, solved with RK45 and calibrated with Bayesian MCMC using PyMC DEMetropolisZ across fourteen free parameters. All five pre-registered hypotheses were confirmed, with posterior super-additivity ranging from thirteen to twenty-two percentage points across model versions. Three sole-authored preprints are currently under review at peer-reviewed journals including IART, PNPBP, and NBR.

The AI-4AI Fellowship provides structured support for graduate study and academic publication, which directly matches my trajectory as an independent computational researcher. I am enrolled in the M.Sc. Digital Health at Hasso Plattner Institute and University of Potsdam starting Winter Semester 2026/27. This programme is a stepping stone to a PhD in computational neuroscience or protein machine learning in the United States or Europe. My existing collaborations with Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU demonstrate readiness for graduate-level research.

My independent output includes a second active research line: TOPOLOGIX, which uses ESM-2 protein-language-model delta-embeddings combined with Morgan/ECFP drug fingerprints and a Random Forest classifier to predict drug-resistance mutations from sequence alone. This system achieves an AUROC of 0.804 plus or minus 0.025 on the Platinum benchmark covering 553 mutations, and 0.634 on SKEMPI 2.0. It beats structure-based baselines such as mCSM-lig at approximately 0.70 while covering one hundred percent of mutations compared to approximately eighteen percent for structure-limited tools.

The fellowship requires a genuine connection between research and artificial intelligence. My CCT model uses Bayesian machine learning for parameter inference. TOPOLOGIX uses transformer-based protein language models. My neurocascade simulation engine couples pharmacokinetics to receptor-binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers, all calibrated with Bayesian methods. These are original computational frameworks built from first principles, not applications of off-the-shelf tools.

I am a Nigerian pharmacist and independent researcher with a B.Pharm from the University of Ibadan, a PCN license, and a German equivalent grade of 1.9. I can commit six to ten hours per week to the fellowship for the full six-month duration. I am open to feedback and committed to producing tangible academic outputs including drafts, proposals, and papers. I will obtain active paid professional membership with the African Institute for Artificial Intelligence before applying.

RESEARCH STATEMENT

My primary research programme for the AI-4AI Fellowship is the Conjunctive Consolidation Threshold model, a tripartite pharmacological framework for preventing reward-memory encoding in addiction. The model consists of three coupled axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. These are formalized as a system of ordinary differential equations solved with RK45 and calibrated using Bayesian Markov Chain Monte Carlo with PyMC DEMetropolisZ across fourteen free parameters. Literature-elicited priors were derived from a systematic screen of 1,847 records. All five pre-registered hypotheses, labeled H1 through H5, were confirmed. Posterior super-additivity ranged from thirteen to twenty-two percentage points across model versions. Three sole-authored preprints are deposited on OSF and Zenodo and are under review at IART, PNPBP, and NBR. A co-authored paper is under review at Alcohol, published by Elsevier.

The artificial intelligence component is central. The Bayesian calibration pipeline is a machine learning method for inferring posterior distributions over parameters that are not directly measurable in human subjects. The model itself generates predictions about which pharmacological interventions, at what doses and timing, can prevent the consolidation of reward memories. This is a testable computational hypothesis that can guide experimental design in animal models and eventually human trials.

A second research line, TOPOLOGIX, directly applies AI to a drug-development problem. Drug-resistance mutations in target proteins render therapies ineffective. Most prediction tools require a protein structure, which is available for only about eighteen percent of clinically relevant mutations. TOPOLOGIX uses ESM-2 protein-language-model delta-embeddings, which capture the effect of a mutation on the protein's internal representation, combined with Morgan circular fingerprints of the drug molecule and a Random Forest classifier. On the Platinum benchmark of 553 mutations, it achieves an AUROC of 0.804 with a standard deviation of 0.025. On SKEMPI 2.0, it achieves 0.634. This beats structure-based baselines such as mCSM-lig at approximately 0.70 while covering all mutations regardless of structure availability.

A third line, neurocascade, is a receptor-to-behavior brain-circuit simulation engine. It couples pharmacokinetics to receptor-binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers. Three literature-calibrated receptor and circuit systems are implemented: mu-opioid, D2 dopamine, and GABA-A. Bayesian calibration was performed with PyMC. Sixty-two of sixty-two tests pass. The circuit-layer parameters are explicitly labeled as illustrative pending real behavioral-data fits.

During the six-month fellowship, I will complete the revision and resubmission of the three CCT model preprints based on journal reviewer feedback, extend TOPOLOGIX to incorporate attention-based interpretability for identifying which sequence positions drive resistance predictions, and begin fitting neurocascade to a published behavioral dataset from the addiction neuroscience literature. These outputs are concrete, measurable, and aligned with the fellowship's emphasis on publication readiness.

SHORT ESSAY: GRADUATE STUDY AMBITION

My graduate study ambition is a PhD in computational neuroscience or protein machine learning at a university in the United States or Europe. The M.Sc. Digital Health at Hasso Plattner Institute and University of Potsdam, starting Winter Semester 2026/27, is the immediate next step. This programme provides formal training in digital health methods, machine learning for biomedical data, and research methodology. It bridges my background as a pharmacist and independent computational researcher with the structured academic environment required for a PhD.

The AI-4AI Fellowship fits directly into this trajectory. The six-month programme provides structured mentorship and a deadline-driven environment for producing academic publications. The three CCT model preprints currently under review will benefit from the feedback and revision process the fellowship supports. The fellowship also connects me with mentors and peers working at the intersection of AI and African research priorities, which is directly relevant to my goal of building a research career that addresses neurological and pharmacological challenges relevant to African populations.

My target regions are the United States, Canada, the United Kingdom, and Europe. I have existing collaborations with neuroscientists at Michigan, Harvard, Princeton, and NYU. These relationships provide a pathway into PhD programmes at those institutions. The fellowship's emphasis on independent work and tangible academic outputs will strengthen my application portfolio.

SHORT ESSAY: INDEPENDENT WORK CAPACITY

I have worked as an independent researcher for two years, from 2024 to the present, producing four active research lines without institutional affiliation or a supervisor. This required self-directed project management, literature review, coding, statistical analysis, and manuscript preparation. I maintain a publication pipeline with three sole-authored preprints under review and a co-authored paper under review at Alcohol.

My technical infrastructure supports independent work. I operate four independent DuckDB-based ingest-to-analyze corpus and RAG pipelines across life sciences, technology and AI and security, and social science domains. I self-host local LLM serving with llama.cpp and on-demand model swapping. I run production systems operations including Linux VPS, systemd, Caddy TLS, CI/CD, and automated backup and disaster-recovery. This infrastructure allows me to run computational experiments, manage literature databases, and produce manuscripts without relying on institutional computing resources.

I can commit six to ten hours per week to the fellowship for the full six-month duration. My current employment as National Product Manager at Synthcare, starting March 2026, is structured to allow dedicated research time outside working hours. I have demonstrated the ability to produce peer-reviewed research outputs while holding full-time employment.

SHORT ESSAY: OPENNESS TO FEEDBACK AND ACADEMIC OUTPUTS

The three CCT model preprints currently under review have already undergone one round of journal review. I responded to each reviewer comment with specific revisions, including adding sensitivity analyses for prior specification, clarifying the biological interpretation of the affective contrast axis, and providing posterior predictive checks for each of the five hypotheses. This process demonstrates my willingness to engage with critical feedback and improve the work accordingly.

During the AI-4AI Fellowship, I commit to producing the following tangible academic outputs: revised and resubmitted manuscripts for all three CCT model preprints, a preprint describing the TOPOLOGIX extension with attention-based interpretability, and a registered report or preprint describing the neurocascade fit to a behavioral dataset. I am open to the mentor redirecting these priorities based on feasibility and alignment with the fellowship's goals.

I also commit to providing regular progress updates, sharing drafts for feedback at specified milestones, and incorporating mentor suggestions into the final outputs. The fellowship's six-month duration is sufficient for one complete revision cycle and one new analysis.

CHECKLIST

- [ ] Obtain active paid professional membership with the African Institute for Artificial Intelligence
- [ ] Prepare CCT model preprint PDFs for submission as writing samples
- [ ] Prepare TOPOLOGIX preprint or code repository link
- [ ] Update ORCID profile with all preprints and publications
- [ ] Request letters of recommendation from Kent Berridge, Samuel Gershman, or Nathaniel Daw
- [ ] Prepare academic transcript from University of Ibadan
- [ ] Prepare proof of enrollment or admission letter from Hasso Plattner Institute
- [ ] Write and proofread motivation letter, research statement, and short essays
- [ ] Submit application through the programme website at careershelpdesk.com
- [ ] Confirm deadline on programme website

EDITOR NOTES

- Eligibility risk: the programme requires active paid professional membership with the African Institute for Artificial Intelligence before applying. Eniola must confirm the membership fee, processing time, and whether it is open to Nigerian residents. This is a mandatory gate condition.
- Facts to verify: the programme URL points to careershelpdesk.com which appears to be a third-party aggregator. Eniola should confirm the official programme page and verify that the fellowship is still accepting applications for 2026. The deadline is listed as "see programme website" which is ambiguous.
- Gap to fill: the profile does not specify Eniola's age or nationality in the application context. The motivation letter should explicitly state Nigerian nationality and age 29 to confirm eligibility for LMIC-track and early-career programmes. This is included in the letter above.
- Gap to fill: the profile does not specify whether Eniola has prior experience with formal mentorship or structured fellowship programmes. If asked in an interview, Eniola should be prepared to describe how they will adapt to a structured programme after working independently.
- Verification needed: the profile lists three preprints as "in review" but does not specify which journals. Eniola should confirm the current status of each submission and have the manuscript IDs or submission confirmation emails ready.

AI Draft — AI-4AI Research Fellowship 2026