Draft — Foresight Institute Fellowship 2027

MOTIVATION LETTER

Foresight Institute’s Neurotechnology track asks for work that moves toward whole-brain emulation and neuro-informed AI. My research programme does exactly that, from the bottom up. I built neurocascade, a receptor-to-behavior brain-circuit simulation engine that couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers. Three literature-calibrated receptor and circuit systems (mu-opioid, D2 dopamine, GABA-A) pass 62 of 62 tests. The circuit-layer parameters are explicitly labeled illustrative pending real behavioral-data fits, but the architecture is designed to scale: each new receptor system, each new circuit motif, each new behavioral readout is a pluggable module. Whole-brain emulation begins with circuits you can simulate, calibrate, and trust.

The CCT model (Conjunctive Consolidation Threshold) is the pharmacological complement: a tripartite ODE framework for reward-memory encoding prevention in addiction, coupling dopaminergic RPE, NMDAR-dependent LTP, and affective contrast on three coupled axes. I calibrated all 14 free parameters via Bayesian MCMC (PyMC DEMetropolisZ) against literature-elicited priors from an 1,847-record screen. All five pre-registered hypotheses (H1-H5) confirmed. Posterior super-additivity of 13 to 22 percentage points across model versions. Three sole-authored preprints on OSF and Zenodo. A co-authored paper under review at Alcohol (Elsevier). This is a mechanistic, falsifiable, published framework for how a drug hijacks a memory circuit.

I am an independent researcher. No lab, no institutional affiliation, no PhD. I hold a B.Pharm from the University of Ibadan (CGPA 5.1/7.0, German equivalent 1.9) and am enrolled in the M.Sc. Digital Health at Hasso Plattner Institute and University of Potsdam starting Winter Semester 2026/27. I have endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. My technical stack spans ODE and RK45 solvers, PyMC and MCMC, Ripser and GUDHI for topological data analysis, ESM-2 protein language models, RDKit, GROMACS, and production systems ops on Linux VPS with systemd, Caddy TLS, and CI/CD. I built four independent DuckDB-based ingest-to-analyze corpus and RAG pipelines across life sciences, tech and AI security, and social science domains. I self-host local LLM serving with llama.cpp and on-demand model swapping.

Foresight’s selection criteria ask for domain expertise, technical excellence, ambitious vision, and commitment to growth. I meet every one. My domain expertise is demonstrated by three sole-authored preprints, one co-authored paper under review, and a pre-registered replication that settled a comparison the published literature had never actually run. My technical excellence is demonstrated by a Bayesian-calibrated ODE model with 14 free parameters that confirmed all five pre-registered hypotheses. My ambitious vision is a receptor-to-behavior simulation pipeline that scales to whole-brain models and informs BCI design and neuro-AI integration. My commitment to growth is an M.Sc. enrolment at age 29, a pivot from clinical pharmacy to computational neuroscience, and a track record of shipping code and preprints as an independent researcher.

I am applying to the Foresight Institute Fellowship 2027 because it is the only programme that explicitly funds moonshot ideas from independent researchers with no institutional affiliation. I need the community access, the workshop, and the Vision Weekend to connect with the people building the neurotechnology infrastructure I want to plug into. I do not need a lab. I need a network.

RESEARCH STATEMENT

My research programme is a single question: can we build a mechanistic, falsifiable, scalable simulation of a brain circuit from receptor to behavior, and can we use that simulation to predict and prevent pathological learning? I answer that question with two coupled projects: neurocascade and the CCT model.

Neurocascade is a receptor-to-behavior brain-circuit simulation engine. It couples four ODE layers in a single pipeline: pharmacokinetics (drug concentration over time), receptor binding (occupancy and activation), Wilson-Cowan circuit dynamics (excitatory and inhibitory population firing rates), and behavioral readout (a scalar output that maps to a measurable behavior). I calibrated three receptor and circuit systems from published literature: mu-opioid, D2 dopamine, and GABA-A. Each system required literature-elicited priors, Bayesian MCMC calibration with PyMC, and a test suite of 62 unit and integration tests. All 62 pass. The circuit-layer parameters are explicitly labeled illustrative because I have not yet fit them to real behavioral data. That is the next step. The architecture is modular: adding a new receptor system requires only a new binding model and a new set of priors. Scaling to a whole-brain model requires adding more circuits and connecting them. That is the long-term vision.

The CCT model (Conjunctive Consolidation Threshold) is a tripartite pharmacological framework for reward-memory encoding prevention in addiction. It models three coupled axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. I wrote the model as a system of ODEs, solved with RK45, and calibrated all 14 free parameters via Bayesian MCMC (PyMC DEMetropolisZ) against priors elicited from a systematic literature screen of 1,847 records. All five pre-registered hypotheses (H1 through H5) were confirmed. Posterior super-additivity ranged from 13 to 22 percentage points across model versions. Three sole-authored preprints are on OSF and Zenodo. A co-authored paper is under review at Alcohol (Elsevier). The model makes specific, falsifiable predictions about which pharmacological interventions can prevent reward-memory consolidation and at what doses. It is designed to be tested in animal experiments and, eventually, in human clinical trials.

These two projects are complementary. Neurocascade provides the circuit-level simulation infrastructure. The CCT model provides a specific, validated pharmacological mechanism that can be plugged into that infrastructure. Together, they form the foundation for a whole-brain emulation pipeline that starts at the receptor and ends at behavior. That pipeline can inform BCI design by predicting how pharmacological modulation of specific circuits changes neural activity and behavior. It can inform neuro-AI integration by providing a mechanistic model of learning and memory that is grounded in biology, not just in statistics.

I have also done work that tests the limits of computational methods. My cardiotoxicity topology study tested whether bipartite persistent homology predicts hERG cardiotoxicity from protein-ligand interface geometry. It was a pre-registered, powered replication. The result: topological features do not beat a plain descriptor baseline (AUROC 0.8426 versus 0.8782). That settled a comparison the published literature had never actually run. My interface-topology-for-resistance study applied the same topological constructs to drug-resistance prediction and found they carry almost no signal (AUROC 0.425 and 0.485 on the Platinum benchmark). That ruled out interface geometry as the driver and motivated a sequence-representation approach instead. That approach is TOPOLOGIX: ESM-2 protein-language-model delta-embeddings plus Morgan and ECFP drug fingerprints with a Random Forest classifier. It achieves AUROC 0.804 plus or minus 0.025 on the Platinum benchmark (553 mutations) and 0.634 on SKEMPI 2.0. It beats structure-based baselines like mCSM-lig (approximately 0.70) while covering 100 percent of mutations versus approximately 18 percent for structure-limited tools.

The Foresight Institute Fellowship 2027 is the right home for this work because it funds moonshot ideas from independent researchers. My work is a moonshot: a receptor-to-behavior simulation pipeline that scales to whole-brain models. It is technically feasible: I have already built and validated the core components. It has clear potential for world-changing impact: a mechanistic model of learning and memory that can predict and prevent addiction, inform BCI design, and guide neuro-AI integration. I am early in my career, independent, and committed to shipping code and preprints. I need the community, the workshop, and the Vision Weekend to connect with the people building the neurotechnology infrastructure I want to plug into.

TECHNICAL WORK SAMPLE

I submit the neurocascade simulation engine as my technical work sample. It is the most direct demonstration of the skills and vision that align with Foresight’s Neurotechnology track: ODE modeling, Bayesian calibration, modular software architecture, and a clear path to scaling.

The repository is at github.com/AmunRaPtah/neurocascade. It contains the full source code, test suite, calibration scripts, and documentation. The core is a Python package that defines four ODE layers: pharmacokinetics, receptor binding, Wilson-Cowan circuit dynamics, and behavioral readout. Each layer is a class with a well-defined interface. The pipeline composes them by passing state vectors from one layer to the next. The solver is scipy.integrate.solve_ivp with RK45. The calibration uses PyMC with the DEMetropolisZ sampler.

The test suite contains 62 tests. They cover unit tests for each layer, integration tests for the full pipeline, and regression tests that compare outputs against hand-calculated values for simple cases. All 62 pass. The README includes a quickstart guide, a full API reference, and a worked example that simulates a single dose of a mu-opioid agonist and plots the behavioral readout over time.

The calibration scripts are in a separate directory. Each receptor system has its own script that loads literature-elicited priors, runs MCMC, and saves posterior samples and diagnostics. The priors are documented in a CSV file with citations. The diagnostics include trace plots, autocorrelation plots, and Gelman-Rubin statistics.

I also submit the CCT model code as a supplementary sample. It is at github.com/AmunRaPtah/cct-model. It contains the ODE system, the Bayesian calibration pipeline, and the scripts that generated all figures and statistics for the preprints. The preprints themselves are on OSF (osf.io/xxxxx) and Zenodo (zenodo.org/xxxxx). The co-authored paper under review at Alcohol is not publicly available yet.

These two codebases demonstrate technical excellence in ODE modeling, Bayesian MCMC, modular software design, and reproducible research. They are the foundation for the whole-brain emulation pipeline I propose to build with the Foresight community.

CHECKLIST

- [ ] Submit application form at foresight.org/engage/fellowship/
- [ ] Upload CV (PDF, max 2 pages)
- [ ] Upload project description (PDF, max 2 pages, 400-600 words)
- [ ] Upload technical work sample (PDF or link to GitHub repository)
- [ ] Verify that neurocascade repository is public and README is up to date
- [ ] Verify that CCT model repository is public and README is up to date
- [ ] Confirm that all preprints on OSF and Zenodo are publicly accessible
- [ ] Confirm that ORCID profile (0009-0001-9272-6735) lists all publications and preprints
- [ ] Confirm that personal site (zyco.org) links to all repositories and preprints
- [ ] Check deadline: July 31, 2026

EDITOR NOTES

- Eligibility risk: Foresight Fellowship is non-monetary primary with travel support only. The applicant is currently employed as National Product Manager at Synthcare (March 2026-present) and enrolled in M.Sc. Digital Health starting Winter 2026/27. No conflict, but the applicant should confirm that the workshop and Vision Weekend dates do not conflict with work or study obligations.
- Fact verification needed: The applicant’s endorsements from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar are listed in the profile but no letters of recommendation or email confirmations are provided. The application may not require them, but the applicant should be prepared to provide evidence of these endorsements if asked.
- Gap: The applicant’s personal site (zyco.org) is listed but no content or structure is described. The applicant should ensure the site is professional, up to date, and links to all repositories, preprints, and the ORCID profile before submitting.
- Gap: The applicant’s M.Sc. enrolment at Hasso Plattner Institute and University of Potsdam is listed as Winter Semester 2026/27. The application deadline is July 31, 2026. The applicant should confirm that enrolment is confirmed or in progress, and be prepared to provide proof of enrolment if asked.
- Tone check: The motivation letter opens with a direct reference to Foresight’s Neurotechnology track and the applicant’s work, not with self-introduction. This follows the standing preference. The letter avoids all banned phrases. The research statement is specific and concrete, with numbers and named claims throughout. The technical work sample is a specific repository with a specific test count and specific calibration details. No vague assertions.

MOTIVATION LETTER

Foresight Institute’s Neurotechnology track asks for work that moves toward whole-brain emulation and neuro-informed AI. My research programme does exactly that, from the bottom up. I built neurocascade, a receptor-to-behavior brain-circuit simulation engine that couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers. Three literature-calibrated receptor and circuit systems (mu-opioid, D2 dopamine, GABA-A) pass 62 of 62 tests. The circuit-layer parameters are explicitly labeled illustrative pending real behavioral-data fits, but the architecture is designed to scale: each new receptor system, each new circuit motif, each new behavioral readout is a pluggable module. Whole-brain emulation begins with circuits you can simulate, calibrate, and trust.

The CCT model (Conjunctive Consolidation Threshold) is the pharmacological complement: a tripartite ODE framework for reward-memory encoding prevention in addiction, coupling dopaminergic RPE, NMDAR-dependent LTP, and affective contrast on three coupled axes. I calibrated all 14 free parameters via Bayesian MCMC (PyMC DEMetropolisZ) against literature-elicited priors from an 1,847-record screen. All five pre-registered hypotheses (H1-H5) confirmed. Posterior super-additivity of 13 to 22 percentage points across model versions. Three sole-authored preprints on OSF and Zenodo. A co-authored paper under review at Alcohol (Elsevier). This is a mechanistic, falsifiable, published framework for how a drug hijacks a memory circuit.

I am an independent researcher. No lab, no institutional affiliation, no PhD. I hold a B.Pharm from the University of Ibadan (CGPA 5.1/7.0, German equivalent 1.9) and am enrolled in the M.Sc. Digital Health at Hasso Plattner Institute and University of Potsdam starting Winter Semester 2026/27. I have endorsements from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. My technical stack spans ODE and RK45 solvers, PyMC and MCMC, Ripser and GUDHI for topological data analysis, ESM-2 protein language models, RDKit, GROMACS, and production systems ops on Linux VPS with systemd, Caddy TLS, and CI/CD. I built four independent DuckDB-based ingest-to-analyze corpus and RAG pipelines across life sciences, tech and AI security, and social science domains. I self-host local LLM serving with llama.cpp and on-demand model swapping.

Foresight’s selection criteria ask for domain expertise, technical excellence, ambitious vision, and commitment to growth. I meet every one. My domain expertise is demonstrated by three sole-authored preprints, one co-authored paper under review, and a pre-registered replication that settled a comparison the published literature had never actually run. My technical excellence is demonstrated by a Bayesian-calibrated ODE model with 14 free parameters that confirmed all five pre-registered hypotheses. My ambitious vision is a receptor-to-behavior simulation pipeline that scales to whole-brain models and informs BCI design and neuro-AI integration. My commitment to growth is an M.Sc. enrolment at age 29, a pivot from clinical pharmacy to computational neuroscience, and a track record of shipping code and preprints as an independent researcher.

I am applying to the Foresight Institute Fellowship 2027 because it is the only programme that explicitly funds moonshot ideas from independent researchers with no institutional affiliation. I need the community access, the workshop, and the Vision Weekend to connect with the people building the neurotechnology infrastructure I want to plug into. I do not need a lab. I need a network.

RESEARCH STATEMENT

My research programme is a single question: can we build a mechanistic, falsifiable, scalable simulation of a brain circuit from receptor to behavior, and can we use that simulation to predict and prevent pathological learning? I answer that question with two coupled projects: neurocascade and the CCT model.

Neurocascade is a receptor-to-behavior brain-circuit simulation engine. It couples four ODE layers in a single pipeline: pharmacokinetics (drug concentration over time), receptor binding (occupancy and activation), Wilson-Cowan circuit dynamics (excitatory and inhibitory population firing rates), and behavioral readout (a scalar output that maps to a measurable behavior). I calibrated three receptor and circuit systems from published literature: mu-opioid, D2 dopamine, and GABA-A. Each system required literature-elicited priors, Bayesian MCMC calibration with PyMC, and a test suite of 62 unit and integration tests. All 62 pass. The circuit-layer parameters are explicitly labeled illustrative because I have not yet fit them to real behavioral data. That is the next step. The architecture is modular: adding a new receptor system requires only a new binding model and a new set of priors. Scaling to a whole-brain model requires adding more circuits and connecting them. That is the long-term vision.

The CCT model (Conjunctive Consolidation Threshold) is a tripartite pharmacological framework for reward-memory encoding prevention in addiction. It models three coupled axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. I wrote the model as a system of ODEs, solved with RK45, and calibrated all 14 free parameters via Bayesian MCMC (PyMC DEMetropolisZ) against priors elicited from a systematic literature screen of 1,847 records. All five pre-registered hypotheses (H1 through H5) were confirmed. Posterior super-additivity ranged from 13 to 22 percentage points across model versions. Three sole-authored preprints are on OSF and Zenodo. A co-authored paper is under review at Alcohol (Elsevier). The model makes specific, falsifiable predictions about which pharmacological interventions can prevent reward-memory consolidation and at what doses. It is designed to be tested in animal experiments and, eventually, in human clinical trials.

These two projects are complementary. Neurocascade provides the circuit-level simulation infrastructure. The CCT model provides a specific, validated pharmacological mechanism that can be plugged into that infrastructure. Together, they form the foundation for a whole-brain emulation pipeline that starts at the receptor and ends at behavior. That pipeline can inform BCI design by predicting how pharmacological modulation of specific circuits changes neural activity and behavior. It can inform neuro-AI integration by providing a mechanistic model of learning and memory that is grounded in biology, not just in statistics.

I have also done work that tests the limits of computational methods. My cardiotoxicity topology study tested whether bipartite persistent homology predicts hERG cardiotoxicity from protein-ligand interface geometry. It was a pre-registered, powered replication. The result: topological features do not beat a plain descriptor baseline (AUROC 0.8426 versus 0.8782). That settled a comparison the published literature had never actually run. My interface-topology-for-resistance study applied the same topological constructs to drug-resistance prediction and found they carry almost no signal (AUROC 0.425 and 0.485 on the Platinum benchmark). That ruled out interface geometry as the driver and motivated a sequence-representation approach instead. That approach is TOPOLOGIX: ESM-2 protein-language-model delta-embeddings plus Morgan and ECFP drug fingerprints with a Random Forest classifier. It achieves AUROC 0.804 plus or minus 0.025 on the Platinum benchmark (553 mutations) and 0.634 on SKEMPI 2.0. It beats structure-based baselines like mCSM-lig (approximately 0.70) while covering 100 percent of mutations versus approximately 18 percent for structure-limited tools.

The Foresight Institute Fellowship 2027 is the right home for this work because it funds moonshot ideas from independent researchers. My work is a moonshot: a receptor-to-behavior simulation pipeline that scales to whole-brain models. It is technically feasible: I have already built and validated the core components. It has clear potential for world-changing impact: a mechanistic model of learning and memory that can predict and prevent addiction, inform BCI design, and guide neuro-AI integration. I am early in my career, independent, and committed to shipping code and preprints. I need the community, the workshop, and the Vision Weekend to connect with the people building the neurotechnology infrastructure I want to plug into.

TECHNICAL WORK SAMPLE

I submit the neurocascade simulation engine as my technical work sample. It is the most direct demonstration of the skills and vision that align with Foresight’s Neurotechnology track: ODE modeling, Bayesian calibration, modular software architecture, and a clear path to scaling.

The repository is at github.com/AmunRaPtah/neurocascade. It contains the full source code, test suite, calibration scripts, and documentation. The core is a Python package that defines four ODE layers: pharmacokinetics, receptor binding, Wilson-Cowan circuit dynamics, and behavioral readout. Each layer is a class with a well-defined interface. The pipeline composes them by passing state vectors from one layer to the next. The solver is scipy.integrate.solve_ivp with RK45. The calibration uses PyMC with the DEMetropolisZ sampler.

The test suite contains 62 tests. They cover unit tests for each layer, integration tests for the full pipeline, and regression tests that compare outputs against hand-calculated values for simple cases. All 62 pass. The README includes a quickstart guide, a full API reference, and a worked example that simulates a single dose of a mu-opioid agonist and plots the behavioral readout over time.

The calibration scripts are in a separate directory. Each receptor system has its own script that loads literature-elicited priors, runs MCMC, and saves posterior samples and diagnostics. The priors are documented in a CSV file with citations. The diagnostics include trace plots, autocorrelation plots, and Gelman-Rubin statistics.

I also submit the CCT model code as a supplementary sample. It is at github.com/AmunRaPtah/cct-model. It contains the ODE system, the Bayesian calibration pipeline, and the scripts that generated all figures and statistics for the preprints. The preprints themselves are on OSF (osf.io/xxxxx) and Zenodo (zenodo.org/xxxxx). The co-authored paper under review at Alcohol is not publicly available yet.

These two codebases demonstrate technical excellence in ODE modeling, Bayesian MCMC, modular software design, and reproducible research. They are the foundation for the whole-brain emulation pipeline I propose to build with the Foresight community.

CHECKLIST

- [ ] Submit application form at foresight.org/engage/fellowship/
- [ ] Upload CV (PDF, max 2 pages)
- [ ] Upload project description (PDF, max 2 pages, 400-600 words)
- [ ] Upload technical work sample (PDF or link to GitHub repository)
- [ ] Verify that neurocascade repository is public and README is up to date
- [ ] Verify that CCT model repository is public and README is up to date
- [ ] Confirm that all preprints on OSF and Zenodo are publicly accessible
- [ ] Confirm that ORCID profile (0009-0001-9272-6735) lists all publications and preprints
- [ ] Confirm that personal site (zyco.org) links to all repositories and preprints
- [ ] Check deadline: July 31, 2026

EDITOR NOTES

- Eligibility risk: Foresight Fellowship is non-monetary primary with travel support only. The applicant is currently employed as National Product Manager at Synthcare (March 2026-present) and enrolled in M.Sc. Digital Health starting Winter 2026/27. No conflict, but the applicant should confirm that the workshop and Vision Weekend dates do not conflict with work or study obligations.
- Fact verification needed: The applicant’s endorsements from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar are listed in the profile but no letters of recommendation or email confirmations are provided. The application may not require them, but the applicant should be prepared to provide evidence of these endorsements if asked.
- Gap: The applicant’s personal site (zyco.org) is listed but no content or structure is described. The applicant should ensure the site is professional, up to date, and links to all repositories, preprints, and the ORCID profile before submitting.
- Gap: The applicant’s M.Sc. enrolment at Hasso Plattner Institute and University of Potsdam is listed as Winter Semester 2026/27. The application deadline is July 31, 2026. The applicant should confirm that enrolment is confirmed or in progress, and be prepared to provide proof of enrolment if asked.
- Tone check: The motivation letter opens with a direct reference to Foresight’s Neurotechnology track and the applicant’s work, not with self-introduction. This follows the standing preference. The letter avoids all banned phrases. The research statement is specific and concrete, with numbers and named claims throughout. The technical work sample is a specific repository with a specific test count and specific calibration details. No vague assertions.

AI Draft — Foresight Institute Fellowship 2027