Draft — Open Philanthropy’s Early-Career Funding for Individuals Interested in Improving the Long-Term Future

MOTIVATION LETTER

The long-term future is shaped by how we understand and manage the systems that govern human behavior and biological risk. My research sits at the intersection of these systems: computational pharmacology, addiction neuroscience, and protein-machine-learning. I am applying to Open Philanthropy's Early-Career Funding for Individuals Interested in Improving the Long-Term Future because my work directly addresses two categories of global catastrophic risk: the societal damage from addiction and the biological threats posed by drug-resistant pathogens.

My primary research line, the Conjunctive Consolidation Threshold (CCT) model, is a tripartite pharmacological framework for preventing reward-memory encoding in addiction. The model couples three axes: dopaminergic reward-prediction error, NMDAR-dependent long-term potentiation, and affective contrast, implemented as a system of ordinary differential equations solved with RK45. I calibrated the model using Bayesian MCMC (PyMC DEMetropolisZ) with 14 free parameters and literature-elicited priors drawn from a systematic screen of 1,847 records. All five pre-registered hypotheses (H1-H5) were confirmed, with posterior super-additivity of 13-22 percentage points across model versions. Three sole-authored preprints are under review at peer-reviewed journals (IART, PNPBP, NBR), and a co-authored paper is under review at Alcohol (Elsevier). This work provides a mechanistic, testable framework for interventions that could prevent addiction before it consolidates, reducing a major source of human suffering and societal instability.

My second research line addresses biological risk directly. The TOPOLOGIX project 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. On the Platinum benchmark (553 mutations), the model achieves AUROC 0.804 with a standard deviation of 0.025; on SKEMPI 2.0, AUROC 0.634. This beats structure-based baselines such as mCSM-lig (approximately 0.70) while covering 100 percent of mutations, compared to roughly 18 percent for structure-limited tools. Predicting resistance mutations from sequence enables faster identification of emerging biological threats and more strong therapeutic design, both critical for biosecurity.

My methodological rigor is demonstrated by a pre-registered, powered replication study on hERG cardiotoxicity. I tested whether bipartite persistent homology predicts cardiotoxicity from protein-ligand interface geometry. The result: topological features do not beat a plain descriptor baseline (AUROC 0.8426 versus 0.8782). This settles a comparison the published literature had never actually run, and it exemplifies my commitment to reporting negative results directly rather than reframing them. The same discipline applies to my ergofluids project, where a Koopman-operator method with a Mori-Zwanzig memory kernel passed synthetic-data gates but failed its first real-data gate; I reported that outcome without reframing.

Open Philanthropy's selection criteria emphasize demonstrated potential for a career contributing to the long-term future, quality of proposal, and need for funding. My track record as an independent researcher, with pre-registered protocols, rigorous Bayesian calibration, and honest reporting, demonstrates that potential. I am enrolled in the M.Sc. Digital Health program at Hasso Plattner Institute / University of Potsdam starting Winter Semester 2026/27, and this funding would support my tuition, research costs, and transition into a full-time research trajectory. My endorsements from Kent Berridge (Michigan), Samuel Gershman (Harvard), Nathaniel Daw (Princeton), and Marcelo Mattar (NYU) attest to the quality of my work.

The funding would enable me to complete the CCT model validation, extend TOPOLOGIX to broader pathogen datasets, and develop the neurocascade simulation engine into a tool for predicting circuit-level responses to pharmacological interventions. These tools contribute directly to reducing global catastrophic risks from addiction and biological threats.

RESEARCH STATEMENT

My research program spans three domains: addiction neuroscience, protein machine learning, and dynamical-systems methods. The unifying thread is the use of computational models to understand and intervene in systems where small changes produce large downstream effects, whether in neural circuits, protein function, or pathogen evolution.

The CCT model addresses addiction as a disorder of memory consolidation. The model posits that reward-memory encoding requires the conjunction of three signals: dopaminergic reward-prediction error, NMDAR-dependent long-term potentiation, and affective contrast. Disrupting any one axis should prevent consolidation. I implemented this as a coupled ODE system (RK45) and calibrated it with Bayesian MCMC using literature-elicited priors from a 1,847-record screen. All five pre-registered hypotheses were confirmed. The model predicts that pharmacological interventions targeting the affective-contrast axis, which is understudied relative to dopamine and glutamate, may offer a novel route to addiction prevention. The next step is to fit the model to behavioral data from animal studies, which requires access to datasets and computational resources that this fellowship would provide.

The TOPOLOGIX project addresses drug resistance as a sequence-encoded phenomenon. I tested whether interface geometry, captured by bipartite persistent homology, predicts resistance. It does not: AUROC 0.425 and 0.485 on the Platinum benchmark. This negative result motivated a shift to sequence representations. ESM-2 delta-embeddings plus Morgan fingerprints and a Random Forest classifier achieve AUROC 0.804 on Platinum and 0.634 on SKEMPI 2.0. The model covers 100 percent of mutations, unlike structure-based tools that cover only 18 percent. This means resistance prediction no longer depends on having a crystal structure, which is critical for emerging pathogens where structures are unavailable. The next phase is to extend the model to antimicrobial resistance in bacterial pathogens, using genomic surveillance data from the GHRU-GSAR project where I worked on AMR genomics.

The neurocascade engine couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics to behavioral readouts. Three systems are calibrated: mu-opioid, D2 dopamine, and GABA-A. All 62 tests pass. The circuit-layer parameters are explicitly labeled illustrative pending real behavioral-data fits. This engine is designed to predict how a drug's receptor profile translates into circuit-level and behavioral effects, which is directly relevant to designing interventions that minimize harm.

My methodological contributions include the honest reporting of negative results. The hERG cardiotoxicity study showed that topological features do not beat a descriptor baseline. The ergofluids project reported a failed real-data gate. These outcomes are published as pre-registered reports, contributing to methodological transparency in computational biology.

The funding from Open Philanthropy would support three specific activities. First, completing the CCT model validation against behavioral data. Second, extending TOPOLOGIX to bacterial AMR prediction. Third, developing the neurocascade engine into a tool for predicting the behavioral effects of novel pharmacological compounds. All three activities contribute to reducing global catastrophic risks: addiction destabilizes societies, and drug resistance undermines modern medicine.

CAREER GOALS ESSAY

My goal is to build a research career at the intersection of computational pharmacology and AI safety, with a focus on reducing global catastrophic risks. I am currently enrolled in the M.Sc. Digital Health program at Hasso Plattner Institute / University of Potsdam, starting Winter Semester 2026/27. This degree provides formal training in digital health methods, complementing my self-directed research experience.

My immediate post-degree goal is to secure a research position, either as a postdoctoral fellow or as a principal investigator at a research institute, where I can continue developing the CCT model and TOPOLOGIX. The CCT model has direct implications for addiction policy and clinical practice; if validated against behavioral data, it could inform the design of prophylactic interventions for individuals at high risk of addiction. TOPOLOGIX has implications for biosecurity; predicting resistance mutations from sequence enables faster identification of emerging threats.

My long-term goal is to establish an independent research group focused on computational models of biological risk. This group would develop tools for predicting how pharmacological interventions affect neural circuits and how pathogens evolve resistance. Both are critical for managing global catastrophic risks in a world where biotechnology is becoming more accessible.

Open Philanthropy's Early-Career Funding is uniquely suited to my situation. I am an independent researcher with a strong publication record but no institutional funding. The flexible funding range of $10K-$100K would cover my tuition at HPI, research computing costs, and conference travel. The rolling deadline allows me to apply now, before the M.Sc. program begins, ensuring continuity of my research.

My endorsements from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar indicate that established researchers in computational neuroscience and reinforcement learning recognize the quality of my work. These connections also provide a pathway to postdoctoral positions after my M.Sc.

The long-term future depends on our ability to model and manage complex biological systems. My research program is designed to build those models, test them rigorously, and make them available to the research community. This funding would accelerate that program at a critical juncture in my career.

PROJECT PROPOSAL: CCT MODEL VALIDATION AND EXTENSION

Project title: Validating the Conjunctive Consolidation Threshold Model Against Behavioral Data and Extending It to Intervention Design

Summary: The CCT model is a tripartite pharmacological framework for reward-memory encoding prevention in addiction. It has been calibrated against literature-derived priors and all five pre-registered hypotheses confirmed. This project will validate the model against behavioral datasets from animal addiction studies and extend it to predict the efficacy of novel pharmacological interventions.

Specific aims:
1. Fit the CCT model to behavioral data from published animal studies of reward-memory consolidation, using Bayesian hierarchical methods to account for between-study variability.
2. Use the fitted model to simulate the effects of interventions targeting each of the three axes (dopaminergic RPE, NMDAR-dependent LTP, affective contrast) and identify the most effective intervention points.
3. Publish the validated model and simulation results as a pre-registered report, following the same standards of transparency I applied to the hERG and ergofluids projects.

Methods: I will extract behavioral data from published studies identified in my existing 1,847-record screen. The ODE model will be extended with a behavioral-readout layer, similar to the neurocascade engine. Bayesian calibration will use PyMC with DEMetropolisZ sampling. Model comparison will use leave-one-study-out cross-validation.

Expected outcomes: A validated model that predicts which pharmacological interventions are most likely to prevent reward-memory consolidation, with quantified uncertainty. This model can inform the design of clinical trials for addiction prevention.

Timeline: Months 1-3, data extraction and preprocessing. Months 4-9, model fitting and validation. Months 10-12, manuscript preparation and submission.

Budget: Tuition support for M.Sc. Digital Health (approximately $15,000), high-performance computing time for MCMC sampling ($5,000), conference travel for presentation at computational neuroscience meetings ($3,000), and publication fees for open-access journals ($2,000). Total: $25,000.

Alignment with Open Philanthropy: This project directly addresses the long-term future by developing a mechanistic understanding of addiction, a major source of societal harm. The methods emphasize transparency and reproducibility, consistent with Open Philanthropy's commitment to rigorous evaluation.

CHECKLIST

- [ ] Complete the online application form at the provided URL
- [ ] Upload motivation letter (this document, 500 words)
- [ ] Upload research statement (this document, 600 words)
- [ ] Upload career goals essay (this document, 350 words)
- [ ] Upload project proposal (this document, 400 words)
- [ ] Request letters of recommendation from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar
- [ ] Upload academic transcripts from University of Ibadan (B.Pharm)
- [ ] Upload proof of enrollment or admission to M.Sc. Digital Health at HPI/Potsdam
- [ ] Upload ORCID record (0009-0001-9272-6735)
- [ ] Upload links to preprints on OSF/Zenodo
- [ ] Verify eligibility for early-career status (age 29, pre-PhD)
- [ ] Confirm rolling deadline and submit as soon as all materials are ready

EDITOR NOTES

- Eligibility risk: The programme targets early-career individuals, typically students or recent graduates. Eniola is 29 and has been out of his B.Pharm since 2021, though he is enrolling in an M.Sc. in 2026/27. The application should emphasize the M.Sc. enrollment as evidence of current early-career status. Verify whether the programme has an age cutoff or time-since-degree limit.
- Verification needed: Confirm the exact name and URL of the programme. The provided URL is a GreaterWrong post, not an official Open Philanthropy application page. The actual application portal may have different questions and word limits. Check the official Open Philanthropy website for the current application form.
- Gap: The profile does not specify how the CCT model's affective-contrast axis is operationalized pharmacologically. If the application requires a technical appendix, Eniola should insert a paragraph specifying the receptor targets and drug classes for this axis.
- Gap: The profile mentions a co-authored paper in Alcohol (Elsevier) under review but does not specify the topic or Eniola's contribution. The application should clarify this to avoid any perception of overclaiming.
- Verification needed: Confirm the German equivalent grade of 1.9 for the B.Pharm CGPA of 5.1/7.0. This conversion should be documented with the official conversion method used by the University of Potsdam or the German grading system.

MOTIVATION LETTER

The long-term future is shaped by how we understand and manage the systems that govern human behavior and biological risk. My research sits at the intersection of these systems: computational pharmacology, addiction neuroscience, and protein-machine-learning. I am applying to Open Philanthropy's Early-Career Funding for Individuals Interested in Improving the Long-Term Future because my work directly addresses two categories of global catastrophic risk: the societal damage from addiction and the biological threats posed by drug-resistant pathogens.

My primary research line, the Conjunctive Consolidation Threshold (CCT) model, is a tripartite pharmacological framework for preventing reward-memory encoding in addiction. The model couples three axes: dopaminergic reward-prediction error, NMDAR-dependent long-term potentiation, and affective contrast, implemented as a system of ordinary differential equations solved with RK45. I calibrated the model using Bayesian MCMC (PyMC DEMetropolisZ) with 14 free parameters and literature-elicited priors drawn from a systematic screen of 1,847 records. All five pre-registered hypotheses (H1-H5) were confirmed, with posterior super-additivity of 13-22 percentage points across model versions. Three sole-authored preprints are under review at peer-reviewed journals (IART, PNPBP, NBR), and a co-authored paper is under review at Alcohol (Elsevier). This work provides a mechanistic, testable framework for interventions that could prevent addiction before it consolidates, reducing a major source of human suffering and societal instability.

My second research line addresses biological risk directly. The TOPOLOGIX project 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. On the Platinum benchmark (553 mutations), the model achieves AUROC 0.804 with a standard deviation of 0.025; on SKEMPI 2.0, AUROC 0.634. This beats structure-based baselines such as mCSM-lig (approximately 0.70) while covering 100 percent of mutations, compared to roughly 18 percent for structure-limited tools. Predicting resistance mutations from sequence enables faster identification of emerging biological threats and more strong therapeutic design, both critical for biosecurity.

My methodological rigor is demonstrated by a pre-registered, powered replication study on hERG cardiotoxicity. I tested whether bipartite persistent homology predicts cardiotoxicity from protein-ligand interface geometry. The result: topological features do not beat a plain descriptor baseline (AUROC 0.8426 versus 0.8782). This settles a comparison the published literature had never actually run, and it exemplifies my commitment to reporting negative results directly rather than reframing them. The same discipline applies to my ergofluids project, where a Koopman-operator method with a Mori-Zwanzig memory kernel passed synthetic-data gates but failed its first real-data gate; I reported that outcome without reframing.

Open Philanthropy's selection criteria emphasize demonstrated potential for a career contributing to the long-term future, quality of proposal, and need for funding. My track record as an independent researcher, with pre-registered protocols, rigorous Bayesian calibration, and honest reporting, demonstrates that potential. I am enrolled in the M.Sc. Digital Health program at Hasso Plattner Institute / University of Potsdam starting Winter Semester 2026/27, and this funding would support my tuition, research costs, and transition into a full-time research trajectory. My endorsements from Kent Berridge (Michigan), Samuel Gershman (Harvard), Nathaniel Daw (Princeton), and Marcelo Mattar (NYU) attest to the quality of my work.

The funding would enable me to complete the CCT model validation, extend TOPOLOGIX to broader pathogen datasets, and develop the neurocascade simulation engine into a tool for predicting circuit-level responses to pharmacological interventions. These tools contribute directly to reducing global catastrophic risks from addiction and biological threats.

RESEARCH STATEMENT

My research program spans three domains: addiction neuroscience, protein machine learning, and dynamical-systems methods. The unifying thread is the use of computational models to understand and intervene in systems where small changes produce large downstream effects, whether in neural circuits, protein function, or pathogen evolution.

The CCT model addresses addiction as a disorder of memory consolidation. The model posits that reward-memory encoding requires the conjunction of three signals: dopaminergic reward-prediction error, NMDAR-dependent long-term potentiation, and affective contrast. Disrupting any one axis should prevent consolidation. I implemented this as a coupled ODE system (RK45) and calibrated it with Bayesian MCMC using literature-elicited priors from a 1,847-record screen. All five pre-registered hypotheses were confirmed. The model predicts that pharmacological interventions targeting the affective-contrast axis, which is understudied relative to dopamine and glutamate, may offer a novel route to addiction prevention. The next step is to fit the model to behavioral data from animal studies, which requires access to datasets and computational resources that this fellowship would provide.

The TOPOLOGIX project addresses drug resistance as a sequence-encoded phenomenon. I tested whether interface geometry, captured by bipartite persistent homology, predicts resistance. It does not: AUROC 0.425 and 0.485 on the Platinum benchmark. This negative result motivated a shift to sequence representations. ESM-2 delta-embeddings plus Morgan fingerprints and a Random Forest classifier achieve AUROC 0.804 on Platinum and 0.634 on SKEMPI 2.0. The model covers 100 percent of mutations, unlike structure-based tools that cover only 18 percent. This means resistance prediction no longer depends on having a crystal structure, which is critical for emerging pathogens where structures are unavailable. The next phase is to extend the model to antimicrobial resistance in bacterial pathogens, using genomic surveillance data from the GHRU-GSAR project where I worked on AMR genomics.

The neurocascade engine couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics to behavioral readouts. Three systems are calibrated: mu-opioid, D2 dopamine, and GABA-A. All 62 tests pass. The circuit-layer parameters are explicitly labeled illustrative pending real behavioral-data fits. This engine is designed to predict how a drug's receptor profile translates into circuit-level and behavioral effects, which is directly relevant to designing interventions that minimize harm.

My methodological contributions include the honest reporting of negative results. The hERG cardiotoxicity study showed that topological features do not beat a descriptor baseline. The ergofluids project reported a failed real-data gate. These outcomes are published as pre-registered reports, contributing to methodological transparency in computational biology.

The funding from Open Philanthropy would support three specific activities. First, completing the CCT model validation against behavioral data. Second, extending TOPOLOGIX to bacterial AMR prediction. Third, developing the neurocascade engine into a tool for predicting the behavioral effects of novel pharmacological compounds. All three activities contribute to reducing global catastrophic risks: addiction destabilizes societies, and drug resistance undermines modern medicine.

CAREER GOALS ESSAY

My goal is to build a research career at the intersection of computational pharmacology and AI safety, with a focus on reducing global catastrophic risks. I am currently enrolled in the M.Sc. Digital Health program at Hasso Plattner Institute / University of Potsdam, starting Winter Semester 2026/27. This degree provides formal training in digital health methods, complementing my self-directed research experience.

My immediate post-degree goal is to secure a research position, either as a postdoctoral fellow or as a principal investigator at a research institute, where I can continue developing the CCT model and TOPOLOGIX. The CCT model has direct implications for addiction policy and clinical practice; if validated against behavioral data, it could inform the design of prophylactic interventions for individuals at high risk of addiction. TOPOLOGIX has implications for biosecurity; predicting resistance mutations from sequence enables faster identification of emerging threats.

My long-term goal is to establish an independent research group focused on computational models of biological risk. This group would develop tools for predicting how pharmacological interventions affect neural circuits and how pathogens evolve resistance. Both are critical for managing global catastrophic risks in a world where biotechnology is becoming more accessible.

Open Philanthropy's Early-Career Funding is uniquely suited to my situation. I am an independent researcher with a strong publication record but no institutional funding. The flexible funding range of $10K-$100K would cover my tuition at HPI, research computing costs, and conference travel. The rolling deadline allows me to apply now, before the M.Sc. program begins, ensuring continuity of my research.

My endorsements from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar indicate that established researchers in computational neuroscience and reinforcement learning recognize the quality of my work. These connections also provide a pathway to postdoctoral positions after my M.Sc.

The long-term future depends on our ability to model and manage complex biological systems. My research program is designed to build those models, test them rigorously, and make them available to the research community. This funding would accelerate that program at a critical juncture in my career.

PROJECT PROPOSAL: CCT MODEL VALIDATION AND EXTENSION

Project title: Validating the Conjunctive Consolidation Threshold Model Against Behavioral Data and Extending It to Intervention Design

Summary: The CCT model is a tripartite pharmacological framework for reward-memory encoding prevention in addiction. It has been calibrated against literature-derived priors and all five pre-registered hypotheses confirmed. This project will validate the model against behavioral datasets from animal addiction studies and extend it to predict the efficacy of novel pharmacological interventions.

Specific aims:
1. Fit the CCT model to behavioral data from published animal studies of reward-memory consolidation, using Bayesian hierarchical methods to account for between-study variability.
2. Use the fitted model to simulate the effects of interventions targeting each of the three axes (dopaminergic RPE, NMDAR-dependent LTP, affective contrast) and identify the most effective intervention points.
3. Publish the validated model and simulation results as a pre-registered report, following the same standards of transparency I applied to the hERG and ergofluids projects.

Methods: I will extract behavioral data from published studies identified in my existing 1,847-record screen. The ODE model will be extended with a behavioral-readout layer, similar to the neurocascade engine. Bayesian calibration will use PyMC with DEMetropolisZ sampling. Model comparison will use leave-one-study-out cross-validation.

Expected outcomes: A validated model that predicts which pharmacological interventions are most likely to prevent reward-memory consolidation, with quantified uncertainty. This model can inform the design of clinical trials for addiction prevention.

Timeline: Months 1-3, data extraction and preprocessing. Months 4-9, model fitting and validation. Months 10-12, manuscript preparation and submission.

Budget: Tuition support for M.Sc. Digital Health (approximately $15,000), high-performance computing time for MCMC sampling ($5,000), conference travel for presentation at computational neuroscience meetings ($3,000), and publication fees for open-access journals ($2,000). Total: $25,000.

Alignment with Open Philanthropy: This project directly addresses the long-term future by developing a mechanistic understanding of addiction, a major source of societal harm. The methods emphasize transparency and reproducibility, consistent with Open Philanthropy's commitment to rigorous evaluation.

CHECKLIST

- [ ] Complete the online application form at the provided URL
- [ ] Upload motivation letter (this document, 500 words)
- [ ] Upload research statement (this document, 600 words)
- [ ] Upload career goals essay (this document, 350 words)
- [ ] Upload project proposal (this document, 400 words)
- [ ] Request letters of recommendation from Kent Berridge, Samuel Gershman, Nathaniel Daw, and Marcelo Mattar
- [ ] Upload academic transcripts from University of Ibadan (B.Pharm)
- [ ] Upload proof of enrollment or admission to M.Sc. Digital Health at HPI/Potsdam
- [ ] Upload ORCID record (0009-0001-9272-6735)
- [ ] Upload links to preprints on OSF/Zenodo
- [ ] Verify eligibility for early-career status (age 29, pre-PhD)
- [ ] Confirm rolling deadline and submit as soon as all materials are ready

EDITOR NOTES

- Eligibility risk: The programme targets early-career individuals, typically students or recent graduates. Eniola is 29 and has been out of his B.Pharm since 2021, though he is enrolling in an M.Sc. in 2026/27. The application should emphasize the M.Sc. enrollment as evidence of current early-career status. Verify whether the programme has an age cutoff or time-since-degree limit.
- Verification needed: Confirm the exact name and URL of the programme. The provided URL is a GreaterWrong post, not an official Open Philanthropy application page. The actual application portal may have different questions and word limits. Check the official Open Philanthropy website for the current application form.
- Gap: The profile does not specify how the CCT model's affective-contrast axis is operationalized pharmacologically. If the application requires a technical appendix, Eniola should insert a paragraph specifying the receptor targets and drug classes for this axis.
- Gap: The profile mentions a co-authored paper in Alcohol (Elsevier) under review but does not specify the topic or Eniola's contribution. The application should clarify this to avoid any perception of overclaiming.
- Verification needed: Confirm the German equivalent grade of 1.9 for the B.Pharm CGPA of 5.1/7.0. This conversion should be documented with the official conversion method used by the University of Potsdam or the German grading system.

AI Draft — Open Philanthropy’s Early-Career Funding for Individuals Interested in Improving the Long-Term Future