Draft — The Evon Labs HealthTech Incubator Program 2026

MOTIVATION LETTER

The Evon Labs HealthTech Incubator Program 2026 targets early-stage founders building technology-driven solutions for low-income and rural populations in Africa. My venture predicts drug resistance mutations from protein sequence alone, without requiring crystal structures, and I am applying as a licensed pharmacist and machine learning engineer with a validated proof-of-concept. The tool achieves an AUROC of 0.804 on the Platinum benchmark across 553 mutations using protein-grouped cross-validation, and it covers 100 percent of mutations versus roughly 18 percent for structure-limited tools. This capability matters most in Nigeria, where tuberculosis, malaria, and antimicrobial resistance are driven by pathogens that mutate faster than laboratories can characterize them.

My background is the regulatory and clinical bridge this programme requires. I am a pharmacist by training, which means I know NAFDAC submission pathways, pharmacovigilance obligations, and the practical realities of drug dispensing in Nigerian community pharmacies. I am also a machine learning engineer who built and validated the model myself. This combination is rare in early-stage healthtech, and it directly addresses the programme's emphasis on navigating regulatory hurdles. The venture is pre-incorporation and pre-revenue, but the proof-of-concept is complete and reproducible. I am applying with a working classifier and a clear path to deployment, not an idea.

The Evon Labs focus on affordable, tech-driven solutions for low-income populations aligns with my deployment model. The model runs on protein sequence data, which is cheap to generate and does not require expensive crystallography infrastructure. A hospital in Ibadan or Kano can sequence a pathogen sample and receive a resistance prediction without shipping samples to a foreign laboratory. This is community diagnostics at scale, and it is why I am targeting this programme specifically rather than a general biotech accelerator.

I am committed to the 10-week intensive hybrid format and have structured my roadmap accordingly. The immediate technical milestone is fine-tuning ESM-2 on the SKEMPI 3K mutation dataset to reach an AUROC of at least 0.70 on that benchmark. The commercial milestone is a pilot with Servier in Suresnes, with whom I have an active conversation. The 30,000 USD grant would fund the compute required for fine-tuning and the legal costs of incorporation in Nigeria. I am asking Evon Labs to accelerate a venture that is already moving, not to start one from zero.

RESEARCH STATEMENT

The venture addresses a specific failure in infectious disease management: drug resistance mutations are identified too slowly and too expensively for low-resource clinical settings. Current state-of-the-art tools like mCSM-lig require protein crystal structures, which exist for only a fraction of clinically relevant mutations. My model removes that dependency entirely. It uses ESM-2 protein language model delta-embeddings combined with ECFP4 drug fingerprints, fed into a Random Forest classifier. The delta-embedding approach captures the change in protein representation caused by a mutation, which is the signal that predicts whether a drug will lose efficacy.

The validation results are concrete. On the Platinum benchmark, which contains 553 mutations and uses protein-grouped cross-validation to prevent data leakage, the model achieves an AUROC of 0.804 with a standard deviation of 0.025. This exceeds the published state-of-the-art performance of mCSM-lig, which sits around 0.70. On SKEMPI 2.0, a harder generalization benchmark, the model achieves 0.634. The coverage advantage is more dramatic: the model predicts on 100 percent of mutations, while structure-limited tools cover approximately 18 percent. In clinical terms, this means the model never returns a null result for lack of structural data.

The technical roadmap has three phases. Phase one, which is the immediate focus, is fine-tuning ESM-2 on the SKEMPI 3K mutation dataset. The goal is to push SKEMPI 2.0 performance from 0.634 to at least 0.70. This requires GPU compute and careful hyperparameter tuning, both of which the Evon Labs grant would fund. Phase two is the Servier pilot in Suresnes, where the model will be tested on their internal resistance datasets. Phase three is the transition to a recurring revenue model, where pharmaceutical companies and diagnostic laboratories license the prediction API on a per-prediction or subscription basis.

The scientific innovation is the integration and the validation discipline, not the individual components, which are all published methods. ESM-2 embeddings have been used for protein function prediction, and ECFP4 fingerprints are standard in cheminformatics, but the delta-embedding representation for resistance prediction, validated with protein-grouped cross-validation to prevent the common pitfall of training and testing on mutations from the same protein, is my contribution. The 0.804 AUROC is the honest output of a rigorous evaluation protocol, not a cherry-picked result.

The deployment context in Nigeria is the differentiator. Tuberculosis resistance in Nigeria is monitored through the National Tuberculosis and Leprosy Control Programme, but whole-genome sequencing capacity is concentrated in Lagos and Abuja. My model changes the economics of resistance testing. A sequence can be generated locally and analyzed against the model without structural biology expertise. This is the affordable, scalable diagnostics pathway that Evon Labs seeks to support.

The venture is at proof-of-concept stage. I am the sole founder and sole author of the codebase. I have not incorporated, and I have no revenue. What I have is a validated classifier, a clear technical roadmap, and a named pilot partner in Servier. The Evon Labs grant would fund the compute, the incorporation, and the first six months of operations.

ESSAY: REGULATORY NAVIGATION

The Evon Labs programme emphasizes demonstrated potential to navigate regulatory hurdles, and this is where my pharmacist license is a structural advantage, not a credential. I trained under the Pharmacists Council of Nigeria curriculum, which includes NAFDAC regulatory affairs, drug registration pathways, and post-market surveillance obligations. I know the difference between a NAFDAC registration for a physical drug product and the regulatory status of a software-as-a-medical-device, which is currently a gray area in Nigeria. This gray area is an opportunity. The venture can engage NAFDAC early, during the pilot phase, rather than after deployment, and shape the conversation around software-based resistance prediction as a diagnostic support tool rather than a standalone medical device.

The regulatory strategy has three layers. First, the model is positioned as a decision-support tool for pharmacists and physicians, not as an autonomous diagnostic. This keeps it within existing professional oversight frameworks. Second, the Servier pilot in France provides a European regulatory reference point, which strengthens the NAFDAC dossier by demonstrating that a major pharmaceutical company has validated the approach. Third, the venture will pursue NAFDAC guidance through the West African Health Organization pathway, which harmonizes regulatory requirements across ECOWAS member states and reduces the cost of multi-country deployment.

The licensed pharmacist identity also matters for the clinical workflow. In Nigerian community pharmacies, the pharmacist is the final checkpoint before a drug is dispensed. A resistance prediction tool that integrates into that workflow, rather than bypassing it, respects the existing professional hierarchy and reduces adoption friction. I am building a tool that makes pharmacists more effective at catching resistance early, which is exactly the kind of regulatory-aware, clinically-grounded innovation the Evon Labs programme is designed to support.

ESSAY: SCALE AND SUSTAINABILITY

The scale story for this venture is not about the Nigerian market alone, but about the entire low-resource diagnostic landscape. The model predicts resistance from protein sequence, which means the marginal cost of a prediction approaches zero once the model is trained. The fixed costs are compute and validation; the variable costs are negligible. This is the economic structure that allows a tool to scale from a single hospital pilot to a national surveillance program without proportional cost increases.

The revenue model has three tiers. Tier one is the pharmaceutical partnership model, exemplified by the Servier pilot, where a pharma company pays for resistance predictions on their proprietary compounds during drug development. This is a high-margin, low-volume revenue stream. Tier two is the diagnostic laboratory model, where labs pay a per-prediction fee for clinical samples. This is a medium-margin, medium-volume stream. Tier three is the public health surveillance model, where ministries of health and international NGOs subscribe to the platform for population-level resistance monitoring. This is a low-margin, high-volume stream that also generates the data needed to improve the model.

The sustainability risk is compute cost. Fine-tuning ESM-2 on SKEMPI 3K requires GPU hours that are expensive in Nigeria due to limited cloud infrastructure. The Evon Labs grant would cover this immediate cost, and the accelerator's compute credits, if available, would extend the runway. The long-term compute strategy is to move inference to CPU-only deployment once the model is fine-tuned, which is feasible because Random Forest inference is computationally light. The heavy compute is a one-time training cost, not a recurring operational cost.

The 10-week programme structure fits my current stage. I am pre-incorporation, which means the programme's mentorship on company formation and intellectual property strategy is as valuable as the grant itself. I have a validated proof-of-concept, which means I can spend the programme on commercialization rather than on basic technical validation. The Evon Labs network in African healthtech would also connect me to the diagnostic laboratory operators and public health officials who are the tier two and tier three customers. I am looking for a programme to accelerate deployment, not to validate an idea.

CHECKLIST

- [ ] Confirm Evon Labs 2026 application deadline from programme website
- [ ] Verify eligibility for pre-incorporation ventures (no explicit restriction found, but confirm)
- [ ] Prepare NAFDAC pharmacist license copy as supporting credential
- [ ] Prepare AUROC validation results summary (Platinum 0.804, SKEMPI 2.0 0.634)
- [ ] Prepare one-page technical appendix describing ESM-2 delta-embedding methodology
- [ ] Secure letter of intent or email confirmation from Servier contact regarding pilot interest
- [ ] Prepare budget breakdown for 30,000 USD grant (compute, incorporation, operations)
- [ ] Confirm commitment to 10-week hybrid programme schedule
- [ ] Prepare pitch deck in Evon Labs required format (check website for template)
- [ ] Identify two references (one technical, one clinical/pharmacy)
- [ ] Draft incorporation plan for Nigeria (CAC registration timeline and cost)
- [ ] Confirm whether programme requires demo video or prototype walkthrough

EDITOR NOTES

- Eligibility risk: the venture is pre-incorporation and pre-revenue, which may be earlier than Evon Labs expects despite the "Idea to Prototype" phase language. The application must emphasize the validated proof-of-concept and named Servier pilot to counter this risk.
- Verification needed: the Servier pilot is described as "in progress" in the profile. The application should not overstate this as a confirmed partnership. The checklist includes securing a letter of intent, which must be done before submission.
- Gap: the profile does not specify the founder's current location or ability to attend a hybrid programme in person. The applicant must confirm whether the 10-week programme requires physical presence at any point and whether travel to the programme location is feasible.
- The framing angle chosen here is the pharma supply chain and community diagnostics positioning, which matches the Evon Labs focus areas. The applicant should not mention the broader AI/biotech venture ambitions in this application; keep the narrative tightly focused on infectious disease resistance in African low-resource settings.
- The profile lists other ventures and research lines (CCT, TOPOLOGIX, neurocascade, ergofluids, psyche-twin), but none of them match the Evon Labs mission. This application correctly uses only the drug resistance prediction venture. Do not cross-reference other ventures in any section.

MOTIVATION LETTER

The Evon Labs HealthTech Incubator Program 2026 targets early-stage founders building technology-driven solutions for low-income and rural populations in Africa. My venture predicts drug resistance mutations from protein sequence alone, without requiring crystal structures, and I am applying as a licensed pharmacist and machine learning engineer with a validated proof-of-concept. The tool achieves an AUROC of 0.804 on the Platinum benchmark across 553 mutations using protein-grouped cross-validation, and it covers 100 percent of mutations versus roughly 18 percent for structure-limited tools. This capability matters most in Nigeria, where tuberculosis, malaria, and antimicrobial resistance are driven by pathogens that mutate faster than laboratories can characterize them.

My background is the regulatory and clinical bridge this programme requires. I am a pharmacist by training, which means I know NAFDAC submission pathways, pharmacovigilance obligations, and the practical realities of drug dispensing in Nigerian community pharmacies. I am also a machine learning engineer who built and validated the model myself. This combination is rare in early-stage healthtech, and it directly addresses the programme's emphasis on navigating regulatory hurdles. The venture is pre-incorporation and pre-revenue, but the proof-of-concept is complete and reproducible. I am applying with a working classifier and a clear path to deployment, not an idea.

The Evon Labs focus on affordable, tech-driven solutions for low-income populations aligns with my deployment model. The model runs on protein sequence data, which is cheap to generate and does not require expensive crystallography infrastructure. A hospital in Ibadan or Kano can sequence a pathogen sample and receive a resistance prediction without shipping samples to a foreign laboratory. This is community diagnostics at scale, and it is why I am targeting this programme specifically rather than a general biotech accelerator.

I am committed to the 10-week intensive hybrid format and have structured my roadmap accordingly. The immediate technical milestone is fine-tuning ESM-2 on the SKEMPI 3K mutation dataset to reach an AUROC of at least 0.70 on that benchmark. The commercial milestone is a pilot with Servier in Suresnes, with whom I have an active conversation. The 30,000 USD grant would fund the compute required for fine-tuning and the legal costs of incorporation in Nigeria. I am asking Evon Labs to accelerate a venture that is already moving, not to start one from zero.

RESEARCH STATEMENT

The venture addresses a specific failure in infectious disease management: drug resistance mutations are identified too slowly and too expensively for low-resource clinical settings. Current state-of-the-art tools like mCSM-lig require protein crystal structures, which exist for only a fraction of clinically relevant mutations. My model removes that dependency entirely. It uses ESM-2 protein language model delta-embeddings combined with ECFP4 drug fingerprints, fed into a Random Forest classifier. The delta-embedding approach captures the change in protein representation caused by a mutation, which is the signal that predicts whether a drug will lose efficacy.

The validation results are concrete. On the Platinum benchmark, which contains 553 mutations and uses protein-grouped cross-validation to prevent data leakage, the model achieves an AUROC of 0.804 with a standard deviation of 0.025. This exceeds the published state-of-the-art performance of mCSM-lig, which sits around 0.70. On SKEMPI 2.0, a harder generalization benchmark, the model achieves 0.634. The coverage advantage is more dramatic: the model predicts on 100 percent of mutations, while structure-limited tools cover approximately 18 percent. In clinical terms, this means the model never returns a null result for lack of structural data.

The technical roadmap has three phases. Phase one, which is the immediate focus, is fine-tuning ESM-2 on the SKEMPI 3K mutation dataset. The goal is to push SKEMPI 2.0 performance from 0.634 to at least 0.70. This requires GPU compute and careful hyperparameter tuning, both of which the Evon Labs grant would fund. Phase two is the Servier pilot in Suresnes, where the model will be tested on their internal resistance datasets. Phase three is the transition to a recurring revenue model, where pharmaceutical companies and diagnostic laboratories license the prediction API on a per-prediction or subscription basis.

The scientific innovation is the integration and the validation discipline, not the individual components, which are all published methods. ESM-2 embeddings have been used for protein function prediction, and ECFP4 fingerprints are standard in cheminformatics, but the delta-embedding representation for resistance prediction, validated with protein-grouped cross-validation to prevent the common pitfall of training and testing on mutations from the same protein, is my contribution. The 0.804 AUROC is the honest output of a rigorous evaluation protocol, not a cherry-picked result.

The deployment context in Nigeria is the differentiator. Tuberculosis resistance in Nigeria is monitored through the National Tuberculosis and Leprosy Control Programme, but whole-genome sequencing capacity is concentrated in Lagos and Abuja. My model changes the economics of resistance testing. A sequence can be generated locally and analyzed against the model without structural biology expertise. This is the affordable, scalable diagnostics pathway that Evon Labs seeks to support.

The venture is at proof-of-concept stage. I am the sole founder and sole author of the codebase. I have not incorporated, and I have no revenue. What I have is a validated classifier, a clear technical roadmap, and a named pilot partner in Servier. The Evon Labs grant would fund the compute, the incorporation, and the first six months of operations.

ESSAY: REGULATORY NAVIGATION

The Evon Labs programme emphasizes demonstrated potential to navigate regulatory hurdles, and this is where my pharmacist license is a structural advantage, not a credential. I trained under the Pharmacists Council of Nigeria curriculum, which includes NAFDAC regulatory affairs, drug registration pathways, and post-market surveillance obligations. I know the difference between a NAFDAC registration for a physical drug product and the regulatory status of a software-as-a-medical-device, which is currently a gray area in Nigeria. This gray area is an opportunity. The venture can engage NAFDAC early, during the pilot phase, rather than after deployment, and shape the conversation around software-based resistance prediction as a diagnostic support tool rather than a standalone medical device.

The regulatory strategy has three layers. First, the model is positioned as a decision-support tool for pharmacists and physicians, not as an autonomous diagnostic. This keeps it within existing professional oversight frameworks. Second, the Servier pilot in France provides a European regulatory reference point, which strengthens the NAFDAC dossier by demonstrating that a major pharmaceutical company has validated the approach. Third, the venture will pursue NAFDAC guidance through the West African Health Organization pathway, which harmonizes regulatory requirements across ECOWAS member states and reduces the cost of multi-country deployment.

The licensed pharmacist identity also matters for the clinical workflow. In Nigerian community pharmacies, the pharmacist is the final checkpoint before a drug is dispensed. A resistance prediction tool that integrates into that workflow, rather than bypassing it, respects the existing professional hierarchy and reduces adoption friction. I am building a tool that makes pharmacists more effective at catching resistance early, which is exactly the kind of regulatory-aware, clinically-grounded innovation the Evon Labs programme is designed to support.

ESSAY: SCALE AND SUSTAINABILITY

The scale story for this venture is not about the Nigerian market alone, but about the entire low-resource diagnostic landscape. The model predicts resistance from protein sequence, which means the marginal cost of a prediction approaches zero once the model is trained. The fixed costs are compute and validation; the variable costs are negligible. This is the economic structure that allows a tool to scale from a single hospital pilot to a national surveillance program without proportional cost increases.

The revenue model has three tiers. Tier one is the pharmaceutical partnership model, exemplified by the Servier pilot, where a pharma company pays for resistance predictions on their proprietary compounds during drug development. This is a high-margin, low-volume revenue stream. Tier two is the diagnostic laboratory model, where labs pay a per-prediction fee for clinical samples. This is a medium-margin, medium-volume stream. Tier three is the public health surveillance model, where ministries of health and international NGOs subscribe to the platform for population-level resistance monitoring. This is a low-margin, high-volume stream that also generates the data needed to improve the model.

The sustainability risk is compute cost. Fine-tuning ESM-2 on SKEMPI 3K requires GPU hours that are expensive in Nigeria due to limited cloud infrastructure. The Evon Labs grant would cover this immediate cost, and the accelerator's compute credits, if available, would extend the runway. The long-term compute strategy is to move inference to CPU-only deployment once the model is fine-tuned, which is feasible because Random Forest inference is computationally light. The heavy compute is a one-time training cost, not a recurring operational cost.

The 10-week programme structure fits my current stage. I am pre-incorporation, which means the programme's mentorship on company formation and intellectual property strategy is as valuable as the grant itself. I have a validated proof-of-concept, which means I can spend the programme on commercialization rather than on basic technical validation. The Evon Labs network in African healthtech would also connect me to the diagnostic laboratory operators and public health officials who are the tier two and tier three customers. I am looking for a programme to accelerate deployment, not to validate an idea.

CHECKLIST

- [ ] Confirm Evon Labs 2026 application deadline from programme website
- [ ] Verify eligibility for pre-incorporation ventures (no explicit restriction found, but confirm)
- [ ] Prepare NAFDAC pharmacist license copy as supporting credential
- [ ] Prepare AUROC validation results summary (Platinum 0.804, SKEMPI 2.0 0.634)
- [ ] Prepare one-page technical appendix describing ESM-2 delta-embedding methodology
- [ ] Secure letter of intent or email confirmation from Servier contact regarding pilot interest
- [ ] Prepare budget breakdown for 30,000 USD grant (compute, incorporation, operations)
- [ ] Confirm commitment to 10-week hybrid programme schedule
- [ ] Prepare pitch deck in Evon Labs required format (check website for template)
- [ ] Identify two references (one technical, one clinical/pharmacy)
- [ ] Draft incorporation plan for Nigeria (CAC registration timeline and cost)
- [ ] Confirm whether programme requires demo video or prototype walkthrough

EDITOR NOTES

- Eligibility risk: the venture is pre-incorporation and pre-revenue, which may be earlier than Evon Labs expects despite the "Idea to Prototype" phase language. The application must emphasize the validated proof-of-concept and named Servier pilot to counter this risk.
- Verification needed: the Servier pilot is described as "in progress" in the profile. The application should not overstate this as a confirmed partnership. The checklist includes securing a letter of intent, which must be done before submission.
- Gap: the profile does not specify the founder's current location or ability to attend a hybrid programme in person. The applicant must confirm whether the 10-week programme requires physical presence at any point and whether travel to the programme location is feasible.
- The framing angle chosen here is the pharma supply chain and community diagnostics positioning, which matches the Evon Labs focus areas. The applicant should not mention the broader AI/biotech venture ambitions in this application; keep the narrative tightly focused on infectious disease resistance in African low-resource settings.
- The profile lists other ventures and research lines (CCT, TOPOLOGIX, neurocascade, ergofluids, psyche-twin), but none of them match the Evon Labs mission. This application correctly uses only the drug resistance prediction venture. Do not cross-reference other ventures in any section.

AI Draft — The Evon Labs HealthTech Incubator Program 2026