Draft — CIFAR Azrieli Global Scholars program

MOTIVATION LETTER

The Conjunctive Consolidation Threshold model, or CCT, is a tripartite pharmacological framework for preventing reward-memory encoding in addiction. It couples three axes, dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast, into a single system of ordinary differential equations solved with RK45 and calibrated through Bayesian MCMC using PyMC's DEMetropolisZ sampler with 14 free parameters. The priors were elicited from a systematic screen of 1,847 records in the literature. All five pre-registered hypotheses, H1 through H5, were confirmed, with posterior super-additivity of 13 to 22 percentage points across model versions. This work is currently under review at three peer-reviewed journals: International Addiction Review, Progress in Neuro-Psychopharmacology and Biological Psychiatry, and Neuroscience and Biobehavioral Reviews.

I am applying to the CIFAR Azrieli Global Scholars program because the CCT model sits precisely at the intersection of two CIFAR research programs: Brain, Mind and Consciousness, and Learning in Machines and Brains. The model is a mechanistic, dynamical-systems explanation of how pharmacological interventions can disrupt the consolidation of reward memories at the circuit level, not a purely descriptive account of addiction. That is the kind of cross-disciplinary problem CIFAR was built to address, one that requires simultaneous fluency in computational neuroscience, receptor pharmacology, and nonlinear dynamics.

The CIFAR network would directly accelerate this research. Kent Berridge at the University of Michigan has published foundational work on incentive salience and reward circuitry, and his framework for affective neuroscience is one of the pillars my model builds upon. Samuel Gershman at Harvard endorsed my recent arXiv submission, and his work on reinforcement learning and memory provides a theoretical grounding for the RPE axis of the CCT model. Nathaniel Daw at Princeton and Marcelo Mattar at NYU have both engaged with the computational assumptions underlying my approach. CIFAR's fellowship structure, which explicitly funds collaboration across its network, would convert these individual touchpoints into sustained, structured engagement.

My leadership experience is unconventional but directly relevant. I run an independent research practice with no institutional safety net. That means every pre-registration, every Bayesian calibration, and every negative result is documented and published openly on OSF and Zenodo. My cardiotoxicity topology study is a case in point. I tested whether bipartite persistent homology could predict hERG cardiotoxicity from protein-ligand interface geometry. The pre-registered, powered replication found that topological features do not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. That negative result settles a comparison the published literature had never actually run. I reported it directly rather than reframing it. That is the standard of rigor I hold myself to, and it is the standard CIFAR fellows are expected to uphold.

The $100,000 CAD in funding would support the next phase of the CCT program: fitting the circuit-layer parameters to real behavioral data, extending the model to relapse scenarios, and building the open-source simulation toolkit that other labs can adopt. The network-building component is equally critical. CIFAR's interdisciplinary mandate matches my conviction that addiction neuroscience cannot advance through pharmacology alone, nor through machine learning alone. It requires both, plus the willingness to publish what does not work.

RESEARCH STATEMENT

The CCT model addresses a specific, unresolved problem in addiction neuroscience: which pharmacological intervention, administered at what time relative to a reward experience, most effectively prevents the encoding of a reward memory that would otherwise drive future drug-seeking behavior. The answer requires a model that integrates three distinct biological scales: the molecular level of receptor binding, the cellular level of synaptic plasticity, and the systems level of affective state.

My approach is to model the interaction of three coupled axes. The first axis is dopaminergic reward prediction error, formalized using temporal difference learning. The second axis is NMDAR-dependent long-term potentiation in the mesolimbic pathway, modeled as a saturating plasticity function. The third axis is affective contrast, the difference between the emotional valence of the drug experience and the baseline affective state. These three axes are coupled into a system of ordinary differential equations, solved with RK45, and calibrated using Bayesian MCMC with PyMC's DEMetropolisZ sampler. The model has 14 free parameters, all with priors elicited from a systematic screen of 1,847 records in the addiction and pharmacology literature.

The model was pre-registered with five hypotheses, H1 through H5, before any calibration was run. All five were confirmed. The central finding is super-additivity: the combined effect of targeting all three axes exceeds the sum of targeting any two, by 13 to 22 percentage points across model versions. This has a direct clinical implication. A single mechanism, such as blocking NMDAR alone or dampening dopaminergic signaling alone, is insufficient. The threshold for memory consolidation is conjunctive, meaning it requires simultaneous modulation of all three axes to fall below the encoding threshold.

The CCT model is currently under review at three peer-reviewed journals: International Addiction Review, Progress in Neuro-Psychopharmacology and Biological Psychiatry, and Neuroscience and Biobehavioral Reviews. The model is also the foundation for neurocascade, a receptor-to-behavior brain-circuit simulation engine I am building. Neurocascade couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers. Three receptor/circuit systems are currently implemented and literature-calibrated: mu-opioid, D2 dopamine, and GABA-A. The engine has 62 passing tests out of 62, and the circuit-layer parameters are explicitly labeled illustrative pending real behavioral-data fits.

The next phase of this research requires access to behavioral datasets that I do not currently have. CIFAR's network includes researchers who collect precisely these data, and the fellowship structure would enable the collaborative fitting of the circuit-layer parameters to real behavioral data. This is the critical validation step that separates a well-calibrated theoretical model from a clinically useful predictive tool.

My broader research program includes TOPOLOGIX, a sequence-based drug-resistance prediction model using ESM-2 protein-language-model delta-embeddings and Morgan fingerprints with a Random Forest classifier. It achieves AUROC 0.804 plus or minus 0.025 on the Platinum benchmark of 553 mutations, beating structure-based baselines such as mCSM-lig at approximately 0.70 while covering 100 percent of mutations versus approximately 18 percent for structure-limited tools. This work is methodologically distinct from CCT but shares the same commitment to pre-registration, open code, and honest reporting of negative results.

CIFAR's Molecular Architecture of Life program is a potential home for the TOPOLOGIX line, but the CCT model is the better fit for this application because it aligns with the Brain, Mind and Consciousness and Learning in Machines and Brains programs, and because it is the research line where CIFAR's collaborative network would have the greatest marginal impact. The model is at the stage where it needs exactly what CIFAR provides: domain experts in affective neuroscience, reinforcement learning theorists, and pharmacologists who can critique and refine the assumptions.

ESSAY: RESEARCH COMPLEMENTARITY

The CCT model complements the Brain, Mind and Consciousness program by offering a mechanistic account of how a specific class of conscious experience, the affective contrast between drug reward and baseline mood, gates memory consolidation. Most computational models of addiction treat reward prediction error as the sole driver of learning. The CCT model formalizes affective contrast as a third, independent axis that interacts multiplicatively with dopaminergic RPE and NMDAR-dependent plasticity. This is a testable, quantitative claim about the role of subjective experience in learning, which is precisely the kind of bridge between computational neuroscience and the study of consciousness that CIFAR's program seeks to build.

The model also complements the Learning in Machines and Brains program by treating the brain as a dynamical system with multiple coupled timescales, rather than as a feedforward function approximator. The Bayesian calibration approach, using literature-elicited priors and posterior sampling, is directly relevant to ongoing work in the program on how biological constraints shape learning algorithms. The finding of super-additivity, where the joint effect of three interventions exceeds the sum of pairwise effects, is a concrete example of a nonlinear interaction that standard machine learning models would miss.

ESSAY: COLLABORATION IDEAS

The most immediate collaboration is with Kent Berridge at the University of Michigan. His work on incentive salience and the role of dopamine in wanting versus liking provides the empirical grounding for the dopaminergic RPE axis of the CCT model. I propose a joint project to test whether the affective contrast axis can be mapped onto the distinction between wanting and liking, using his lab's behavioral paradigms to generate the data needed to fit the circuit-layer parameters of neurocascade.

A second collaboration is with Samuel Gershman at Harvard. His work on reinforcement learning and memory, particularly on the role of predictive representations, could refine the RPE axis of the CCT model. I propose a theoretical collaboration to extend the model to relapse scenarios, where the reward memory is partially consolidated and must be disrupted rather than prevented.

A third collaboration is with Marcelo Mattar at NYU, whose work on memory consolidation and replay could inform the temporal dynamics of the CCT model. The model currently assumes a fixed consolidation window; Mattar's framework could make that window a function of replay events, making the model more biologically realistic.

ESSAY: LEADERSHIP EXPERIENCE

My leadership is demonstrated through the independent operation of a multi-domain computational research practice. I have designed and executed five distinct research lines since 2024, each with pre-registered protocols, open code repositories, and published results on OSF and Zenodo. This requires project management, technical skill across Python, R, HPC environments, and the discipline to report negative results without reframing them.

I also lead in open-science practice. My cardiotoxicity topology study is the first to run the comparison between topological features and descriptor baselines that the literature had claimed but never actually executed. The result, that topology does not beat the baseline, is now public knowledge. That is leadership by example: showing that a well-designed negative result is a contribution.

My employment history includes a National Product Manager role at Synthcare, where I manage product strategy across a national portfolio, and prior roles as a clinical pharmacist and bioinformatics researcher. These roles required coordinating across teams with different technical vocabularies, a skill directly transferable to CIFAR's interdisciplinary collaborations.

ESSAY: ENGAGEMENT WITH NON-ACADEMIC COMMUNITIES

As a licensed pharmacist in Nigeria, I have direct experience translating research findings into clinical practice. My work at Ramset Pharmacy and Synthcare involves daily interaction with patients and clinicians, which grounds my research in real-world constraints. The CCT model's clinical implication, that addiction treatment requires conjunctive rather than single-mechanism interventions, is directly relevant to the Nigerian context, where opioid and methamphetamine use are growing public health concerns.

I also maintain a public-facing research presence through my personal site, zyco.org, and my GitHub repository, where all code and pre-registrations are publicly available. I have built four independent data pipelines for corpus ingestion and analysis across life sciences, tech and AI, and social science domains, and I self-host local LLM serving infrastructure. This technical infrastructure is not a research output in itself, but it enables transparent, reproducible research that non-academic audiences can inspect and reuse.

CHECKLIST

- [ ] Verify current eligibility status for CIFAR Azrieli Global Scholars, specifically the PhD or equivalent degree requirement and the five-year window for full-time academic appointment
- [ ] Confirm whether the B.Pharm degree plus enrollment in the M.Sc. Digital Health program satisfies the eligibility criteria, or whether an exception applies
- [ ] Secure three letters of recommendation, at least one from a faculty member or equivalent, preferably one from a non-academic leader
- [ ] Confirm that Kent Berridge, Samuel Gershman, Nathaniel Daw, or Marcelo Mattar is willing to serve as a recommender or provide a supporting statement
- [ ] Verify the current review status of the three CCT preprints at IART, PNPBP, and NBR, and update the application with any acceptances
- [ ] Prepare the full CCT model documentation, including the pre-registration, the 1,847-record screen methodology, and the posterior distributions for all 14 parameters
- [ ] Prepare the neurocascade repository with the 62 passing tests and the labeled illustrative circuit-layer parameters
- [ ] Confirm the exact word limits for each essay question on the CIFAR application portal, as the limits may differ from the fallback assumptions used here
- [ ] Verify the application deadline on the official CIFAR website, as the URL provided is a secondary source
- [ ] Draft a one-page CV formatted to CIFAR specifications, emphasizing the CCT model, the negative result on hERG topology, and the open-science record
- [ ] Prepare a budget breakdown for the $100,000 CAD, specifying the portion allocated to data acquisition for behavioral fits, compute resources, and conference travel

EDITOR NOTES

- Eligibility risk: the CIFAR Azrieli Global Scholars program typically requires a PhD or equivalent degree and a full-time academic appointment within the first five years. The applicant has a B.Pharm, is enrolled in an M.Sc., and works as an independent researcher. This is a significant eligibility risk that must be verified before investing time in the full application.
- The recommendation letters are the single most important unknown. The profile lists endorsements and collaborations with Berridge, Gershman, Daw, and Mattar, but it is not confirmed that any of them will write a letter. This must be secured early.
- The CCT model is the correct research line to lead with, but the application should not claim that the circuit-layer parameters of neurocascade are validated against real behavioral data. They are explicitly labeled illustrative. The essay on collaboration ideas should be written as proposals, not as confirmed projects.
- The applicant's employment timeline shows a National Product Manager role at Synthcare starting March 2026. This is a full-time industry role. The application must clarify how this role coexists with the independent research program and the M.Sc. enrollment, as CIFAR may view this as a conflict with the full-time academic appointment requirement.

MOTIVATION LETTER

The Conjunctive Consolidation Threshold model, or CCT, is a tripartite pharmacological framework for preventing reward-memory encoding in addiction. It couples three axes, dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast, into a single system of ordinary differential equations solved with RK45 and calibrated through Bayesian MCMC using PyMC's DEMetropolisZ sampler with 14 free parameters. The priors were elicited from a systematic screen of 1,847 records in the literature. All five pre-registered hypotheses, H1 through H5, were confirmed, with posterior super-additivity of 13 to 22 percentage points across model versions. This work is currently under review at three peer-reviewed journals: International Addiction Review, Progress in Neuro-Psychopharmacology and Biological Psychiatry, and Neuroscience and Biobehavioral Reviews.

I am applying to the CIFAR Azrieli Global Scholars program because the CCT model sits precisely at the intersection of two CIFAR research programs: Brain, Mind and Consciousness, and Learning in Machines and Brains. The model is a mechanistic, dynamical-systems explanation of how pharmacological interventions can disrupt the consolidation of reward memories at the circuit level, not a purely descriptive account of addiction. That is the kind of cross-disciplinary problem CIFAR was built to address, one that requires simultaneous fluency in computational neuroscience, receptor pharmacology, and nonlinear dynamics.

The CIFAR network would directly accelerate this research. Kent Berridge at the University of Michigan has published foundational work on incentive salience and reward circuitry, and his framework for affective neuroscience is one of the pillars my model builds upon. Samuel Gershman at Harvard endorsed my recent arXiv submission, and his work on reinforcement learning and memory provides a theoretical grounding for the RPE axis of the CCT model. Nathaniel Daw at Princeton and Marcelo Mattar at NYU have both engaged with the computational assumptions underlying my approach. CIFAR's fellowship structure, which explicitly funds collaboration across its network, would convert these individual touchpoints into sustained, structured engagement.

My leadership experience is unconventional but directly relevant. I run an independent research practice with no institutional safety net. That means every pre-registration, every Bayesian calibration, and every negative result is documented and published openly on OSF and Zenodo. My cardiotoxicity topology study is a case in point. I tested whether bipartite persistent homology could predict hERG cardiotoxicity from protein-ligand interface geometry. The pre-registered, powered replication found that topological features do not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. That negative result settles a comparison the published literature had never actually run. I reported it directly rather than reframing it. That is the standard of rigor I hold myself to, and it is the standard CIFAR fellows are expected to uphold.

The $100,000 CAD in funding would support the next phase of the CCT program: fitting the circuit-layer parameters to real behavioral data, extending the model to relapse scenarios, and building the open-source simulation toolkit that other labs can adopt. The network-building component is equally critical. CIFAR's interdisciplinary mandate matches my conviction that addiction neuroscience cannot advance through pharmacology alone, nor through machine learning alone. It requires both, plus the willingness to publish what does not work.

RESEARCH STATEMENT

The CCT model addresses a specific, unresolved problem in addiction neuroscience: which pharmacological intervention, administered at what time relative to a reward experience, most effectively prevents the encoding of a reward memory that would otherwise drive future drug-seeking behavior. The answer requires a model that integrates three distinct biological scales: the molecular level of receptor binding, the cellular level of synaptic plasticity, and the systems level of affective state.

My approach is to model the interaction of three coupled axes. The first axis is dopaminergic reward prediction error, formalized using temporal difference learning. The second axis is NMDAR-dependent long-term potentiation in the mesolimbic pathway, modeled as a saturating plasticity function. The third axis is affective contrast, the difference between the emotional valence of the drug experience and the baseline affective state. These three axes are coupled into a system of ordinary differential equations, solved with RK45, and calibrated using Bayesian MCMC with PyMC's DEMetropolisZ sampler. The model has 14 free parameters, all with priors elicited from a systematic screen of 1,847 records in the addiction and pharmacology literature.

The model was pre-registered with five hypotheses, H1 through H5, before any calibration was run. All five were confirmed. The central finding is super-additivity: the combined effect of targeting all three axes exceeds the sum of targeting any two, by 13 to 22 percentage points across model versions. This has a direct clinical implication. A single mechanism, such as blocking NMDAR alone or dampening dopaminergic signaling alone, is insufficient. The threshold for memory consolidation is conjunctive, meaning it requires simultaneous modulation of all three axes to fall below the encoding threshold.

The CCT model is currently under review at three peer-reviewed journals: International Addiction Review, Progress in Neuro-Psychopharmacology and Biological Psychiatry, and Neuroscience and Biobehavioral Reviews. The model is also the foundation for neurocascade, a receptor-to-behavior brain-circuit simulation engine I am building. Neurocascade couples pharmacokinetics to receptor binding to Wilson-Cowan circuit dynamics to behavioral-readout ODE layers. Three receptor/circuit systems are currently implemented and literature-calibrated: mu-opioid, D2 dopamine, and GABA-A. The engine has 62 passing tests out of 62, and the circuit-layer parameters are explicitly labeled illustrative pending real behavioral-data fits.

The next phase of this research requires access to behavioral datasets that I do not currently have. CIFAR's network includes researchers who collect precisely these data, and the fellowship structure would enable the collaborative fitting of the circuit-layer parameters to real behavioral data. This is the critical validation step that separates a well-calibrated theoretical model from a clinically useful predictive tool.

My broader research program includes TOPOLOGIX, a sequence-based drug-resistance prediction model using ESM-2 protein-language-model delta-embeddings and Morgan fingerprints with a Random Forest classifier. It achieves AUROC 0.804 plus or minus 0.025 on the Platinum benchmark of 553 mutations, beating structure-based baselines such as mCSM-lig at approximately 0.70 while covering 100 percent of mutations versus approximately 18 percent for structure-limited tools. This work is methodologically distinct from CCT but shares the same commitment to pre-registration, open code, and honest reporting of negative results.

CIFAR's Molecular Architecture of Life program is a potential home for the TOPOLOGIX line, but the CCT model is the better fit for this application because it aligns with the Brain, Mind and Consciousness and Learning in Machines and Brains programs, and because it is the research line where CIFAR's collaborative network would have the greatest marginal impact. The model is at the stage where it needs exactly what CIFAR provides: domain experts in affective neuroscience, reinforcement learning theorists, and pharmacologists who can critique and refine the assumptions.

ESSAY: RESEARCH COMPLEMENTARITY

The CCT model complements the Brain, Mind and Consciousness program by offering a mechanistic account of how a specific class of conscious experience, the affective contrast between drug reward and baseline mood, gates memory consolidation. Most computational models of addiction treat reward prediction error as the sole driver of learning. The CCT model formalizes affective contrast as a third, independent axis that interacts multiplicatively with dopaminergic RPE and NMDAR-dependent plasticity. This is a testable, quantitative claim about the role of subjective experience in learning, which is precisely the kind of bridge between computational neuroscience and the study of consciousness that CIFAR's program seeks to build.

The model also complements the Learning in Machines and Brains program by treating the brain as a dynamical system with multiple coupled timescales, rather than as a feedforward function approximator. The Bayesian calibration approach, using literature-elicited priors and posterior sampling, is directly relevant to ongoing work in the program on how biological constraints shape learning algorithms. The finding of super-additivity, where the joint effect of three interventions exceeds the sum of pairwise effects, is a concrete example of a nonlinear interaction that standard machine learning models would miss.

ESSAY: COLLABORATION IDEAS

The most immediate collaboration is with Kent Berridge at the University of Michigan. His work on incentive salience and the role of dopamine in wanting versus liking provides the empirical grounding for the dopaminergic RPE axis of the CCT model. I propose a joint project to test whether the affective contrast axis can be mapped onto the distinction between wanting and liking, using his lab's behavioral paradigms to generate the data needed to fit the circuit-layer parameters of neurocascade.

A second collaboration is with Samuel Gershman at Harvard. His work on reinforcement learning and memory, particularly on the role of predictive representations, could refine the RPE axis of the CCT model. I propose a theoretical collaboration to extend the model to relapse scenarios, where the reward memory is partially consolidated and must be disrupted rather than prevented.

A third collaboration is with Marcelo Mattar at NYU, whose work on memory consolidation and replay could inform the temporal dynamics of the CCT model. The model currently assumes a fixed consolidation window; Mattar's framework could make that window a function of replay events, making the model more biologically realistic.

ESSAY: LEADERSHIP EXPERIENCE

My leadership is demonstrated through the independent operation of a multi-domain computational research practice. I have designed and executed five distinct research lines since 2024, each with pre-registered protocols, open code repositories, and published results on OSF and Zenodo. This requires project management, technical skill across Python, R, HPC environments, and the discipline to report negative results without reframing them.

I also lead in open-science practice. My cardiotoxicity topology study is the first to run the comparison between topological features and descriptor baselines that the literature had claimed but never actually executed. The result, that topology does not beat the baseline, is now public knowledge. That is leadership by example: showing that a well-designed negative result is a contribution.

My employment history includes a National Product Manager role at Synthcare, where I manage product strategy across a national portfolio, and prior roles as a clinical pharmacist and bioinformatics researcher. These roles required coordinating across teams with different technical vocabularies, a skill directly transferable to CIFAR's interdisciplinary collaborations.

ESSAY: ENGAGEMENT WITH NON-ACADEMIC COMMUNITIES

As a licensed pharmacist in Nigeria, I have direct experience translating research findings into clinical practice. My work at Ramset Pharmacy and Synthcare involves daily interaction with patients and clinicians, which grounds my research in real-world constraints. The CCT model's clinical implication, that addiction treatment requires conjunctive rather than single-mechanism interventions, is directly relevant to the Nigerian context, where opioid and methamphetamine use are growing public health concerns.

I also maintain a public-facing research presence through my personal site, zyco.org, and my GitHub repository, where all code and pre-registrations are publicly available. I have built four independent data pipelines for corpus ingestion and analysis across life sciences, tech and AI, and social science domains, and I self-host local LLM serving infrastructure. This technical infrastructure is not a research output in itself, but it enables transparent, reproducible research that non-academic audiences can inspect and reuse.

CHECKLIST

- [ ] Verify current eligibility status for CIFAR Azrieli Global Scholars, specifically the PhD or equivalent degree requirement and the five-year window for full-time academic appointment
- [ ] Confirm whether the B.Pharm degree plus enrollment in the M.Sc. Digital Health program satisfies the eligibility criteria, or whether an exception applies
- [ ] Secure three letters of recommendation, at least one from a faculty member or equivalent, preferably one from a non-academic leader
- [ ] Confirm that Kent Berridge, Samuel Gershman, Nathaniel Daw, or Marcelo Mattar is willing to serve as a recommender or provide a supporting statement
- [ ] Verify the current review status of the three CCT preprints at IART, PNPBP, and NBR, and update the application with any acceptances
- [ ] Prepare the full CCT model documentation, including the pre-registration, the 1,847-record screen methodology, and the posterior distributions for all 14 parameters
- [ ] Prepare the neurocascade repository with the 62 passing tests and the labeled illustrative circuit-layer parameters
- [ ] Confirm the exact word limits for each essay question on the CIFAR application portal, as the limits may differ from the fallback assumptions used here
- [ ] Verify the application deadline on the official CIFAR website, as the URL provided is a secondary source
- [ ] Draft a one-page CV formatted to CIFAR specifications, emphasizing the CCT model, the negative result on hERG topology, and the open-science record
- [ ] Prepare a budget breakdown for the $100,000 CAD, specifying the portion allocated to data acquisition for behavioral fits, compute resources, and conference travel

EDITOR NOTES

- Eligibility risk: the CIFAR Azrieli Global Scholars program typically requires a PhD or equivalent degree and a full-time academic appointment within the first five years. The applicant has a B.Pharm, is enrolled in an M.Sc., and works as an independent researcher. This is a significant eligibility risk that must be verified before investing time in the full application.
- The recommendation letters are the single most important unknown. The profile lists endorsements and collaborations with Berridge, Gershman, Daw, and Mattar, but it is not confirmed that any of them will write a letter. This must be secured early.
- The CCT model is the correct research line to lead with, but the application should not claim that the circuit-layer parameters of neurocascade are validated against real behavioral data. They are explicitly labeled illustrative. The essay on collaboration ideas should be written as proposals, not as confirmed projects.
- The applicant's employment timeline shows a National Product Manager role at Synthcare starting March 2026. This is a full-time industry role. The application must clarify how this role coexists with the independent research program and the M.Sc. enrollment, as CIFAR may view this as a conflict with the full-time academic appointment requirement.

AI Draft — CIFAR Azrieli Global Scholars program