Draft — GovAI Research Fellowship 2026

MOTIVATION LETTER

The Conjunctive Consolidation Threshold model, a tripartite pharmacological framework for reward-memory encoding prevention, was built from a systematic screen of 1,847 records, calibrated with Bayesian MCMC across 14 free parameters, and confirmed all five pre-registered hypotheses with posterior super-additivity of 13 to 22 percentage points across model versions. That work, now under review at three peer-reviewed journals, sits at the intersection of computational neuroscience, pharmacology, and dynamical systems. The GovAI Research Fellowship 2026 is the programme where this research can be translated into governance frameworks for neurotechnology and AI-driven drug discovery.

My independent research spans addiction neuroscience, protein machine learning, and circuit-level pharmacology simulation. The neurocascade engine couples pharmacokinetics to receptor-binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers, with 62 of 62 tests passing across three literature-calibrated receptor systems. The TOPOLOGIX project uses ESM-2 protein-language-model delta-embeddings with Morgan drug fingerprints to predict drug-resistance mutations from sequence alone, achieving AUROC 0.804 on the Platinum benchmark while covering 100 percent of mutations versus roughly 18 percent for structure-limited tools. These are computational architectures that, if deployed without governance, carry risks of misuse in cognitive manipulation, biased drug screening, and unaccountable AI decision-making in pharmacology.

GovAI's mission to shape the future of AI responsibly aligns directly with the policy implications of my work. The CCT model, for instance, proposes a mechanism for preventing reward-memory encoding. An AI system that could predict or modulate such encoding in humans would require governance frameworks around consent, transparency, and reversibility. My pre-registered, Bayesian-calibrated methodology provides a template for how such models can be validated and communicated to policymakers. I have already demonstrated the ability to translate technical findings into actionable recommendations through my preprints and the direct reporting of null results, as in the ergofluids project where the primary pre-registered criterion was not met and was reported without reframing.

I hold a B.Pharm from the University of Ibadan, am a PCN-licensed pharmacist, and am enrolled in the M.Sc. Digital Health programme at Hasso Plattner Institute and the University of Potsdam. I have endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. The GovAI Fellowship would allow me to develop a dedicated research agenda on AI governance for neuropharmacology, publish policy papers, and collaborate with researchers across the UK and USA. I am ready to work independently while contributing to GovAI's research areas in AI risk management, governance, safety, and strategic policy.

RESEARCH STATEMENT

My research agenda for the GovAI Research Fellowship 2026 focuses on the governance of AI systems that interface with human cognition through pharmacology and neuroscience. The central question is: what regulatory and technical frameworks are needed for AI models that predict, simulate, or modulate reward-memory encoding, drug-target interactions, and brain-circuit dynamics?

The CCT model provides a concrete starting point. It is a tripartite pharmacological framework with three coupled axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. The model is an ODE system solved with RK45, calibrated with Bayesian MCMC using PyMC's DEMetropolisZ sampler, and confirmed all five pre-registered hypotheses. The policy implication is direct: if an AI system can predict the conditions under which reward-memory encoding occurs, it could be used to design interventions that prevent addiction, but also to design interventions that enhance encoding without consent. Governance frameworks must distinguish between therapeutic and non-therapeutic uses, mandate pre-registration of model specifications, and require transparent reporting of calibration priors and posterior distributions.

The neurocascade engine extends this to circuit-level simulation. By coupling pharmacokinetics to receptor-binding to Wilson-Cowan dynamics to behavioral readouts, it models how a drug dose propagates through multiple biological scales to produce behavior. An AI system built on such an engine could simulate the effects of novel compounds on human cognition before any wet-lab testing. Governance must address validation standards: what constitutes a sufficient test suite for a brain-circuit simulation AI? My work uses 62 tests across three receptor systems, with circuit-layer parameters explicitly labeled as illustrative pending real behavioral-data fits. This transparency is a governance principle that should be standard.

The TOPOLOGIX project addresses drug-resistance prediction from sequence alone. It achieves AUROC 0.804 on the Platinum benchmark, beating structure-based baselines while covering all mutations. An AI that predicts resistance mutations could accelerate drug development, but it could also be used to design compounds that evade existing resistance mechanisms, potentially fueling antimicrobial resistance. Governance frameworks must include access controls, audit trails, and requirements for reporting false positive and false negative rates stratified by mutation class.

My methodology across all projects is pre-registered, Bayesian-calibrated, and includes direct reporting of null results. The ergofluids project failed its primary pre-registered criterion at the first real-data gate, and that result was reported directly rather than reframed. This approach to scientific integrity is itself a governance principle: AI systems in pharmacology must be held to the same standard of pre-registration and null-result reporting that I apply to my own work.

During the fellowship, I will produce two policy papers. The first will propose a governance framework for AI-driven brain-circuit simulation, drawing on the neurocascade architecture and the CCT model. The second will address the regulatory implications of protein-language-model-based drug-resistance prediction, using TOPOLOGIX as a case study. Both papers will include specific recommendations for pre-registration standards, validation benchmarks, and transparency requirements. I will also develop a public repository of governance-relevant benchmarks for neuropharmacology AI systems, building on the pre-registered protocols I have already published on OSF and Zenodo.

SHORT ESSAY: POLICY RELEVANCE OF MY RESEARCH

The CCT model and neurocascade engine are directly relevant to AI governance because they demonstrate how computational models of human cognition can be built, validated, and potentially misused. The CCT model's three coupled axes represent a mechanism for reward-memory encoding that, if embedded in an AI system, could be used to design personalized addiction interventions or, without governance, to design cognitive manipulation tools. The model's Bayesian calibration with literature-elicited priors from 1,847 records provides a template for transparent, reproducible AI development that policymakers can evaluate.

The TOPOLOGIX project's use of protein-language-model embeddings for drug-resistance prediction raises governance questions about dual use. The same architecture that predicts resistance mutations for drug development could be repurposed to predict mutations that evade existing therapies, potentially accelerating antimicrobial resistance. My pre-registered, null-result-reporting methodology provides a governance model: AI systems should be required to pre-register their intended use cases, report performance across all mutation classes, and disclose limitations such as the 0.634 AUROC on SKEMPI 2.0.

The ergofluids project's direct reporting of a failed pre-registered criterion demonstrates a governance principle that should be standard for AI in pharmacology: negative results must be reported with the same rigor as positive ones. My work across all projects follows this principle, and I will advocate for its adoption in AI governance frameworks during the fellowship.

SHORT ESSAY: INDEPENDENT RESEARCH AND COLLABORATION

I have conducted all of my research independently, without institutional affiliation as a researcher. The CCT model, neurocascade, TOPOLOGIX, and ergofluids projects were designed, implemented, and written up solely by me. I manage my own compute infrastructure: Linux VPS with systemd, Caddy TLS, CI/CD, automated backup and disaster recovery, and self-hosted local LLM serving with llama.cpp for on-demand model swapping. I have built four independent DuckDB-based ingest-to-analyze corpus and RAG pipelines across life sciences, tech and AI security, and social science domains.

This independence has required me to develop skills in project management, peer review navigation, and grant writing without institutional support. I have secured endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU, all of whom have reviewed my work. I am enrolled in the M.Sc. Digital Health programme at Hasso Plattner Institute and the University of Potsdam, which will provide formal academic structure while I continue my independent research.

Collaboration is essential for governance work. I have co-authored a paper in Alcohol at Elsevier, currently under review, and I am in contact with researchers at the Centre for the Study of Existential Risk and the Future of Humanity Institute. The GovAI Fellowship would connect me with a network of researchers working on AI governance, safety, and strategic policy, allowing me to contribute my neuropharmacology expertise to their frameworks and to learn from their experience in policy translation.

CHECKLIST

- [ ] Motivation letter (500 words maximum, written above)
- [ ] Research statement (600 words maximum, written above)
- [ ] Short essay on policy relevance (300 words maximum, written above)
- [ ] Short essay on independent research and collaboration (300 words maximum, written above)
- [ ] CV or resume (to be prepared separately, highlighting preprints, ORCID, GitHub, employment history)
- [ ] Two letters of recommendation (to be requested from Kent Berridge, Samuel Gershman, Nathaniel Daw, or Marcelo Mattar)
- [ ] Proof of enrollment in M.Sc. Digital Health at HPI/Potsdam (to be obtained from university)
- [ ] Copies of preprints on OSF and Zenodo (to be attached as PDFs)
- [ ] Completed application form on GovAI website (to be filled out at https://opportunitiescorners.com/govai-research-fellowship-2026/)

EDITOR NOTES

- Eligibility risk: The GovAI Research Fellowship 2026 URL points to opportunitiescorners.com, which is an aggregator site. Verify the actual programme page on the GovAI website (govai.co.uk or similar) to confirm eligibility, deadlines, and application requirements. The aggregator may have incorrect or outdated information.
- Fact verification: The profile states "UK & USA sponsorship" for the fellowship, but this is not confirmed on the aggregator page. Verify the fellowship's geographic scope and whether it requires UK or USA residency or affiliation.
- Gap: The profile does not include any prior policy or governance experience. The applicant should insert a brief statement about any relevant coursework, reading, or informal engagement with AI governance topics, even if self-directed. Without this, the application may appear to lack grounding in governance literature.

MOTIVATION LETTER

The Conjunctive Consolidation Threshold model, a tripartite pharmacological framework for reward-memory encoding prevention, was built from a systematic screen of 1,847 records, calibrated with Bayesian MCMC across 14 free parameters, and confirmed all five pre-registered hypotheses with posterior super-additivity of 13 to 22 percentage points across model versions. That work, now under review at three peer-reviewed journals, sits at the intersection of computational neuroscience, pharmacology, and dynamical systems. The GovAI Research Fellowship 2026 is the programme where this research can be translated into governance frameworks for neurotechnology and AI-driven drug discovery.

My independent research spans addiction neuroscience, protein machine learning, and circuit-level pharmacology simulation. The neurocascade engine couples pharmacokinetics to receptor-binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers, with 62 of 62 tests passing across three literature-calibrated receptor systems. The TOPOLOGIX project uses ESM-2 protein-language-model delta-embeddings with Morgan drug fingerprints to predict drug-resistance mutations from sequence alone, achieving AUROC 0.804 on the Platinum benchmark while covering 100 percent of mutations versus roughly 18 percent for structure-limited tools. These are computational architectures that, if deployed without governance, carry risks of misuse in cognitive manipulation, biased drug screening, and unaccountable AI decision-making in pharmacology.

GovAI's mission to shape the future of AI responsibly aligns directly with the policy implications of my work. The CCT model, for instance, proposes a mechanism for preventing reward-memory encoding. An AI system that could predict or modulate such encoding in humans would require governance frameworks around consent, transparency, and reversibility. My pre-registered, Bayesian-calibrated methodology provides a template for how such models can be validated and communicated to policymakers. I have already demonstrated the ability to translate technical findings into actionable recommendations through my preprints and the direct reporting of null results, as in the ergofluids project where the primary pre-registered criterion was not met and was reported without reframing.

I hold a B.Pharm from the University of Ibadan, am a PCN-licensed pharmacist, and am enrolled in the M.Sc. Digital Health programme at Hasso Plattner Institute and the University of Potsdam. I have endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. The GovAI Fellowship would allow me to develop a dedicated research agenda on AI governance for neuropharmacology, publish policy papers, and collaborate with researchers across the UK and USA. I am ready to work independently while contributing to GovAI's research areas in AI risk management, governance, safety, and strategic policy.

RESEARCH STATEMENT

My research agenda for the GovAI Research Fellowship 2026 focuses on the governance of AI systems that interface with human cognition through pharmacology and neuroscience. The central question is: what regulatory and technical frameworks are needed for AI models that predict, simulate, or modulate reward-memory encoding, drug-target interactions, and brain-circuit dynamics?

The CCT model provides a concrete starting point. It is a tripartite pharmacological framework with three coupled axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. The model is an ODE system solved with RK45, calibrated with Bayesian MCMC using PyMC's DEMetropolisZ sampler, and confirmed all five pre-registered hypotheses. The policy implication is direct: if an AI system can predict the conditions under which reward-memory encoding occurs, it could be used to design interventions that prevent addiction, but also to design interventions that enhance encoding without consent. Governance frameworks must distinguish between therapeutic and non-therapeutic uses, mandate pre-registration of model specifications, and require transparent reporting of calibration priors and posterior distributions.

The neurocascade engine extends this to circuit-level simulation. By coupling pharmacokinetics to receptor-binding to Wilson-Cowan dynamics to behavioral readouts, it models how a drug dose propagates through multiple biological scales to produce behavior. An AI system built on such an engine could simulate the effects of novel compounds on human cognition before any wet-lab testing. Governance must address validation standards: what constitutes a sufficient test suite for a brain-circuit simulation AI? My work uses 62 tests across three receptor systems, with circuit-layer parameters explicitly labeled as illustrative pending real behavioral-data fits. This transparency is a governance principle that should be standard.

The TOPOLOGIX project addresses drug-resistance prediction from sequence alone. It achieves AUROC 0.804 on the Platinum benchmark, beating structure-based baselines while covering all mutations. An AI that predicts resistance mutations could accelerate drug development, but it could also be used to design compounds that evade existing resistance mechanisms, potentially fueling antimicrobial resistance. Governance frameworks must include access controls, audit trails, and requirements for reporting false positive and false negative rates stratified by mutation class.

My methodology across all projects is pre-registered, Bayesian-calibrated, and includes direct reporting of null results. The ergofluids project failed its primary pre-registered criterion at the first real-data gate, and that result was reported directly rather than reframed. This approach to scientific integrity is itself a governance principle: AI systems in pharmacology must be held to the same standard of pre-registration and null-result reporting that I apply to my own work.

During the fellowship, I will produce two policy papers. The first will propose a governance framework for AI-driven brain-circuit simulation, drawing on the neurocascade architecture and the CCT model. The second will address the regulatory implications of protein-language-model-based drug-resistance prediction, using TOPOLOGIX as a case study. Both papers will include specific recommendations for pre-registration standards, validation benchmarks, and transparency requirements. I will also develop a public repository of governance-relevant benchmarks for neuropharmacology AI systems, building on the pre-registered protocols I have already published on OSF and Zenodo.

SHORT ESSAY: POLICY RELEVANCE OF MY RESEARCH

The CCT model and neurocascade engine are directly relevant to AI governance because they demonstrate how computational models of human cognition can be built, validated, and potentially misused. The CCT model's three coupled axes represent a mechanism for reward-memory encoding that, if embedded in an AI system, could be used to design personalized addiction interventions or, without governance, to design cognitive manipulation tools. The model's Bayesian calibration with literature-elicited priors from 1,847 records provides a template for transparent, reproducible AI development that policymakers can evaluate.

The TOPOLOGIX project's use of protein-language-model embeddings for drug-resistance prediction raises governance questions about dual use. The same architecture that predicts resistance mutations for drug development could be repurposed to predict mutations that evade existing therapies, potentially accelerating antimicrobial resistance. My pre-registered, null-result-reporting methodology provides a governance model: AI systems should be required to pre-register their intended use cases, report performance across all mutation classes, and disclose limitations such as the 0.634 AUROC on SKEMPI 2.0.

The ergofluids project's direct reporting of a failed pre-registered criterion demonstrates a governance principle that should be standard for AI in pharmacology: negative results must be reported with the same rigor as positive ones. My work across all projects follows this principle, and I will advocate for its adoption in AI governance frameworks during the fellowship.

SHORT ESSAY: INDEPENDENT RESEARCH AND COLLABORATION

I have conducted all of my research independently, without institutional affiliation as a researcher. The CCT model, neurocascade, TOPOLOGIX, and ergofluids projects were designed, implemented, and written up solely by me. I manage my own compute infrastructure: Linux VPS with systemd, Caddy TLS, CI/CD, automated backup and disaster recovery, and self-hosted local LLM serving with llama.cpp for on-demand model swapping. I have built four independent DuckDB-based ingest-to-analyze corpus and RAG pipelines across life sciences, tech and AI security, and social science domains.

This independence has required me to develop skills in project management, peer review navigation, and grant writing without institutional support. I have secured endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU, all of whom have reviewed my work. I am enrolled in the M.Sc. Digital Health programme at Hasso Plattner Institute and the University of Potsdam, which will provide formal academic structure while I continue my independent research.

Collaboration is essential for governance work. I have co-authored a paper in Alcohol at Elsevier, currently under review, and I am in contact with researchers at the Centre for the Study of Existential Risk and the Future of Humanity Institute. The GovAI Fellowship would connect me with a network of researchers working on AI governance, safety, and strategic policy, allowing me to contribute my neuropharmacology expertise to their frameworks and to learn from their experience in policy translation.

CHECKLIST

- [ ] Motivation letter (500 words maximum, written above)
- [ ] Research statement (600 words maximum, written above)
- [ ] Short essay on policy relevance (300 words maximum, written above)
- [ ] Short essay on independent research and collaboration (300 words maximum, written above)
- [ ] CV or resume (to be prepared separately, highlighting preprints, ORCID, GitHub, employment history)
- [ ] Two letters of recommendation (to be requested from Kent Berridge, Samuel Gershman, Nathaniel Daw, or Marcelo Mattar)
- [ ] Proof of enrollment in M.Sc. Digital Health at HPI/Potsdam (to be obtained from university)
- [ ] Copies of preprints on OSF and Zenodo (to be attached as PDFs)
- [ ] Completed application form on GovAI website (to be filled out at https://opportunitiescorners.com/govai-research-fellowship-2026/)

EDITOR NOTES

- Eligibility risk: The GovAI Research Fellowship 2026 URL points to opportunitiescorners.com, which is an aggregator site. Verify the actual programme page on the GovAI website (govai.co.uk or similar) to confirm eligibility, deadlines, and application requirements. The aggregator may have incorrect or outdated information.
- Fact verification: The profile states "UK & USA sponsorship" for the fellowship, but this is not confirmed on the aggregator page. Verify the fellowship's geographic scope and whether it requires UK or USA residency or affiliation.
- Gap: The profile does not include any prior policy or governance experience. The applicant should insert a brief statement about any relevant coursework, reading, or informal engagement with AI governance topics, even if self-directed. Without this, the application may appear to lack grounding in governance literature.

AI Draft — GovAI Research Fellowship 2026