Draft — lvlup.vc/post/what-investors-look-for-before-the-metrics

MOTIVATION LETTER

The Conjunctive Consolidation Threshold model predicts an 85.8 percent reduction in reward-memory encoding probability. I built that model, validated it with ODE/RK45 and Bayesian MCMC, and filed a provisional patent on its core architecture in Q3 2026. I did all of this as an independent researcher in Lagos, Nigeria, with no institutional funding, no PhD supervisor, and no lab. The lvlup.vc programme asks what investors look for before the metrics. The answer is execution under constraint, and that is exactly what I have demonstrated.

I am a licensed pharmacist who spent years dispensing naltrexone and acamprosate to patients who relapsed within weeks. That clinical frustration drove me to ask a different question: what if addiction is a failure of the brain's conjunctive consolidation threshold rather than a failure of will? I formalised that question into a tripartite pharmacological framework, wrote three sole-authored preprints, and submitted a review article to Neuroscience and Biobehavioral Reviews. My model is endorsed by Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. These endorsements confirm a working mathematical architecture.

I also ship products. IMPRINT screens addiction liability. TOPOLOGIX uses topological data analysis with persistent homology and bipartite simplicial complexes to predict drug-protein interactions, with a hERG cardiotoxicity MVP already running. GATE evaluates BCI neural-stimulation safety under Apache 2.0. Each platform was built alone, in Lagos, on a laptop, with open-source tools. I learned TDA from scratch because I needed it. I learned Bayesian population dynamics because the model demanded it. I learned Nextflow and SLURM because the simulations required HPC.

The lvlup.vc programme values founders who have lived the problem, who execute quickly, and who balance confidence with curiosity. I have lived the problem as a pharmacist. I have executed by publishing, patenting, and building three platforms in under eighteen months. And I remain curious: I am applying to the MSc programme at MUG in Graz, Austria, starting October 2026, because I need formal training in computational neuroscience to push the CCT model into clinical trial design. I am not chasing trends. I am solving a specific, measurable problem: preventing reward-memory consolidation in addiction.

SHORT ESSAY: PROBLEM INSIGHT

Addiction is a memory disorder. Every relapse is a memory retrieval event: the cue, the context, the expectation of reward. My clinical experience at Ramset Pharmacy in Lagos confirmed this. Patients knew the consequences. They wanted to stop. But the memory trace of the drug experience had been consolidated below a threshold that made it resistant to extinction.

The Conjunctive Consolidation Threshold model formalises this. It posits that reward-memory encoding requires simultaneous activation of three pathways: dopaminergic salience, glutamatergic plasticity, and opioidergic hedonic tone. If any pathway falls below a conjunctive threshold, consolidation fails. My simulations show that triple-pharmacological blockade reduces encoding probability from 0.855 to 0.122, a reduction of 85.8 percent, with super-additivity of 12.8 percentage points beyond additive predictions. All five pre-registered hypotheses H1 through H5 were confirmed.

This is not a theoretical exercise. I have filed a provisional patent on the CCT core architecture. I have built IMPRINT to screen compounds for addiction liability. I am designing a Bayesian adaptive clinical trial architecture that can test the model in human subjects. The problem insight came from watching patients fail. The solution came from mathematics, pharmacology, and computational modelling.

SHORT ESSAY: EXECUTION AND RESOURCEFULNESS

I built three platforms with no funding. IMPRINT runs on Supabase and Postgres with a Python backend. TOPOLOGIX uses Ripser and Gudhi for persistent homology, RDKit for molecular descriptors, and scikit-learn for classification. GATE is a BCI safety evaluation tool written in Python with NEURON and Brian2 for neural simulation, released under Apache 2.0. Each platform was built in Lagos, on consumer hardware, during evenings and weekends while working as a clinical pharmacist and later as National Product Manager at Synthcare.

The CCT model required ODE integration with RK45, Bayesian MCMC with PyMC, and population dynamics simulations that I ran on a rented cloud instance because my laptop could not handle the matrix sizes. I learned Nextflow and SLURM to parallelise the runs. I learned TDA because I needed to analyse drug-protein interaction topologies for TOPOLOGIX. I learned AlphaFold and AutoDock for the NMDA/insulin docking work at CDDDP.

Resourcefulness is not a buzzword for me. It is the only option. I have no lab, no grant, no supervisor. I have a laptop, an ORCID, a GitHub account, and a provisional patent. That is enough to validate a mathematical model, build three platforms, and get endorsements from four of the world's leading computational neuroscientists.

SHORT ESSAY: CONVICTION AND ADAPTABILITY

Conviction means filing a patent before you have a PhD. It means submitting a review article to Neuroscience and Biobehavioral Reviews as an independent researcher from Nigeria. It means cold-emailing Kent Berridge and Samuel Gershman and getting responses because the model was rigorous enough to warrant their time.

Adaptability means listening. When Nathaniel Daw pointed out a weakness in my temporal difference learning assumptions, I rewrote the formal specification. When Marcelo Mattar suggested Bayesian population dynamics, I learned PyMC and ran the MCMC chains. When the reviewers at Alcohol asked for additional validation on the co-authored paper, I ran the extra simulations within a week.

I am applying to the MSc programme at MUG in Graz because I need formal training in computational neuroscience. I am applying to lvlup.vc because I need the strategic guidance and network to turn a validated model into a clinical trial. I am not attached to any particular path. I am attached to the problem: preventing reward-memory consolidation in addiction. Everything else is iteration.

CHECKLIST

- [ ] Motivation letter (300-500 words, first person, no markdown)
- [ ] Short essay on problem insight (200-350 words)
- [ ] Short essay on execution and resourcefulness (200-350 words)
- [ ] Short essay on conviction and adaptability (200-350 words)
- [ ] CV or resume (attach as PDF)
- [ ] Links to preprints: OSF 10.17605/OSF.IO/KG7B5, OSF 10.17605/OSF.IO/EMY4U, Zenodo 10.5281/zenodo.20492472
- [ ] Link to GitHub: github.com/AmunRaPtah
- [ ] Link to ORCID: 0009-0001-9272-6735
- [ ] Link to zyco.org
- [ ] Provisional patent filing confirmation (Q3 2026)
- [ ] Endorsement letters or contact details for Kent Berridge, Samuel Gershman, Nathaniel Daw, Marcelo Mattar
- [ ] Proof of B.Pharm degree and PCN license
- [ ] Proof of employment at Synthcare and Ramset Pharmacy

EDITOR NOTES

- Eligibility risk: lvlup.vc appears to be a venture capital blog post, not a fellowship programme. Verify whether this is an actual application or a reading assignment. If it is a reading assignment, the applicant may need to write a reflection or pitch deck instead of a motivation letter.
- Fact verification: confirm the provisional patent filing date and jurisdiction. The profile says Q3 2026, which is in the future relative to the current date. If the application is submitted before Q3 2026, rephrase as "provisional patent application filed" or "provisional patent in preparation."
- Gap: the applicant's age (29) and nationality (Nigerian) are strong signals for LMIC-track programmes, but lvlup.vc may not have an LMIC track. Confirm whether the programme has geographic or career-stage restrictions. If not, the motivation letter should still emphasise the Nigeria angle as a demonstration of resourcefulness.
- Gap: the applicant is not yet enrolled in an MSc programme. If lvlup.vc requires current enrolment in a degree programme, this may be a disqualifier. The applicant should confirm eligibility before submitting.
- Gap: the applicant's employment as National Product Manager at Synthcare started in March 2026. If the application is submitted before that date, adjust the timeline or omit the role.

MOTIVATION LETTER

The Conjunctive Consolidation Threshold model predicts an 85.8 percent reduction in reward-memory encoding probability. I built that model, validated it with ODE/RK45 and Bayesian MCMC, and filed a provisional patent on its core architecture in Q3 2026. I did all of this as an independent researcher in Lagos, Nigeria, with no institutional funding, no PhD supervisor, and no lab. The lvlup.vc programme asks what investors look for before the metrics. The answer is execution under constraint, and that is exactly what I have demonstrated.

I am a licensed pharmacist who spent years dispensing naltrexone and acamprosate to patients who relapsed within weeks. That clinical frustration drove me to ask a different question: what if addiction is a failure of the brain's conjunctive consolidation threshold rather than a failure of will? I formalised that question into a tripartite pharmacological framework, wrote three sole-authored preprints, and submitted a review article to Neuroscience and Biobehavioral Reviews. My model is endorsed by Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU. These endorsements confirm a working mathematical architecture.

I also ship products. IMPRINT screens addiction liability. TOPOLOGIX uses topological data analysis with persistent homology and bipartite simplicial complexes to predict drug-protein interactions, with a hERG cardiotoxicity MVP already running. GATE evaluates BCI neural-stimulation safety under Apache 2.0. Each platform was built alone, in Lagos, on a laptop, with open-source tools. I learned TDA from scratch because I needed it. I learned Bayesian population dynamics because the model demanded it. I learned Nextflow and SLURM because the simulations required HPC.

The lvlup.vc programme values founders who have lived the problem, who execute quickly, and who balance confidence with curiosity. I have lived the problem as a pharmacist. I have executed by publishing, patenting, and building three platforms in under eighteen months. And I remain curious: I am applying to the MSc programme at MUG in Graz, Austria, starting October 2026, because I need formal training in computational neuroscience to push the CCT model into clinical trial design. I am not chasing trends. I am solving a specific, measurable problem: preventing reward-memory consolidation in addiction.

SHORT ESSAY: PROBLEM INSIGHT

Addiction is a memory disorder. Every relapse is a memory retrieval event: the cue, the context, the expectation of reward. My clinical experience at Ramset Pharmacy in Lagos confirmed this. Patients knew the consequences. They wanted to stop. But the memory trace of the drug experience had been consolidated below a threshold that made it resistant to extinction.

The Conjunctive Consolidation Threshold model formalises this. It posits that reward-memory encoding requires simultaneous activation of three pathways: dopaminergic salience, glutamatergic plasticity, and opioidergic hedonic tone. If any pathway falls below a conjunctive threshold, consolidation fails. My simulations show that triple-pharmacological blockade reduces encoding probability from 0.855 to 0.122, a reduction of 85.8 percent, with super-additivity of 12.8 percentage points beyond additive predictions. All five pre-registered hypotheses H1 through H5 were confirmed.

This is not a theoretical exercise. I have filed a provisional patent on the CCT core architecture. I have built IMPRINT to screen compounds for addiction liability. I am designing a Bayesian adaptive clinical trial architecture that can test the model in human subjects. The problem insight came from watching patients fail. The solution came from mathematics, pharmacology, and computational modelling.

SHORT ESSAY: EXECUTION AND RESOURCEFULNESS

I built three platforms with no funding. IMPRINT runs on Supabase and Postgres with a Python backend. TOPOLOGIX uses Ripser and Gudhi for persistent homology, RDKit for molecular descriptors, and scikit-learn for classification. GATE is a BCI safety evaluation tool written in Python with NEURON and Brian2 for neural simulation, released under Apache 2.0. Each platform was built in Lagos, on consumer hardware, during evenings and weekends while working as a clinical pharmacist and later as National Product Manager at Synthcare.

The CCT model required ODE integration with RK45, Bayesian MCMC with PyMC, and population dynamics simulations that I ran on a rented cloud instance because my laptop could not handle the matrix sizes. I learned Nextflow and SLURM to parallelise the runs. I learned TDA because I needed to analyse drug-protein interaction topologies for TOPOLOGIX. I learned AlphaFold and AutoDock for the NMDA/insulin docking work at CDDDP.

Resourcefulness is not a buzzword for me. It is the only option. I have no lab, no grant, no supervisor. I have a laptop, an ORCID, a GitHub account, and a provisional patent. That is enough to validate a mathematical model, build three platforms, and get endorsements from four of the world's leading computational neuroscientists.

SHORT ESSAY: CONVICTION AND ADAPTABILITY

Conviction means filing a patent before you have a PhD. It means submitting a review article to Neuroscience and Biobehavioral Reviews as an independent researcher from Nigeria. It means cold-emailing Kent Berridge and Samuel Gershman and getting responses because the model was rigorous enough to warrant their time.

Adaptability means listening. When Nathaniel Daw pointed out a weakness in my temporal difference learning assumptions, I rewrote the formal specification. When Marcelo Mattar suggested Bayesian population dynamics, I learned PyMC and ran the MCMC chains. When the reviewers at Alcohol asked for additional validation on the co-authored paper, I ran the extra simulations within a week.

I am applying to the MSc programme at MUG in Graz because I need formal training in computational neuroscience. I am applying to lvlup.vc because I need the strategic guidance and network to turn a validated model into a clinical trial. I am not attached to any particular path. I am attached to the problem: preventing reward-memory consolidation in addiction. Everything else is iteration.

CHECKLIST

- [ ] Motivation letter (300-500 words, first person, no markdown)
- [ ] Short essay on problem insight (200-350 words)
- [ ] Short essay on execution and resourcefulness (200-350 words)
- [ ] Short essay on conviction and adaptability (200-350 words)
- [ ] CV or resume (attach as PDF)
- [ ] Links to preprints: OSF 10.17605/OSF.IO/KG7B5, OSF 10.17605/OSF.IO/EMY4U, Zenodo 10.5281/zenodo.20492472
- [ ] Link to GitHub: github.com/AmunRaPtah
- [ ] Link to ORCID: 0009-0001-9272-6735
- [ ] Link to zyco.org
- [ ] Provisional patent filing confirmation (Q3 2026)
- [ ] Endorsement letters or contact details for Kent Berridge, Samuel Gershman, Nathaniel Daw, Marcelo Mattar
- [ ] Proof of B.Pharm degree and PCN license
- [ ] Proof of employment at Synthcare and Ramset Pharmacy

EDITOR NOTES

- Eligibility risk: lvlup.vc appears to be a venture capital blog post, not a fellowship programme. Verify whether this is an actual application or a reading assignment. If it is a reading assignment, the applicant may need to write a reflection or pitch deck instead of a motivation letter.
- Fact verification: confirm the provisional patent filing date and jurisdiction. The profile says Q3 2026, which is in the future relative to the current date. If the application is submitted before Q3 2026, rephrase as "provisional patent application filed" or "provisional patent in preparation."
- Gap: the applicant's age (29) and nationality (Nigerian) are strong signals for LMIC-track programmes, but lvlup.vc may not have an LMIC track. Confirm whether the programme has geographic or career-stage restrictions. If not, the motivation letter should still emphasise the Nigeria angle as a demonstration of resourcefulness.
- Gap: the applicant is not yet enrolled in an MSc programme. If lvlup.vc requires current enrolment in a degree programme, this may be a disqualifier. The applicant should confirm eligibility before submitting.
- Gap: the applicant's employment as National Product Manager at Synthcare started in March 2026. If the application is submitted before that date, adjust the timeline or omit the role.

AI Draft — lvlup.vc/post/what-investors-look-for-before-the-metrics