MOTIVATION LETTER The Platinum benchmark contains 553 mutations across 53 proteins. My model predicts drug resistance mutations from protein sequence alone with an AUROC of 0.804, plus or minus 0.025, under protein-grouped cross-validation. The published state of the art, mCSM-lig, scores 0.70 and requires a crystal structure. My model covers 100 percent of mutations in the benchmark; structure-limited tools cover roughly 18 percent. This is the gap I bring to HEC Paris Incubateur. My name is Eniola Olutogun. I trained as a pharmacist, then as a machine learning engineer. I built this venture alone, from the ESM-2 protein language model delta-embeddings and ECFP4 drug fingerprints through the Random Forest classifier and the validation runs. The science is proven. What I need now is the business discipline and investor network to turn a validated proof of concept into a company with revenue. HEC Paris Incubateur selects 18 to 24 projects per batch. The jury evaluates founder profile, business model viability, market potential, and fit with the program's values: ambitious, executor, collaborative, knowledgeable, humble, unique. I meet those criteria. I am the sole author of a model that beats published state of the art. I have named pilot discussions with Servier in Suresnes and a research partnership track at Paris-Saclay I2BC and Institut Pasteur. I have applied to IncubAlliance and AI House, and I am in active conversation with SEMIA and Quest for Health. What I lack is a structured environment to refine pricing, packaging, and go-to-market strategy. That is precisely what HEC's four-month program provides. The commercial path is concrete. The roadmap is to fine-tune ESM-2 on the SKEMPI 3K mutation set, reach an AUROC of at least 0.70 on that expanded benchmark, then convert the Servier pilot into an annual recurring revenue contract. The target market is pharma AI and computational drug discovery, specifically oncology, antivirals, and antimicrobial resistance. Every major pharma company runs resistance screening during lead optimization. Most of them are blind to mutations that require a crystal structure to assess. My model removes that constraint. I am not asking HEC to validate my science. I am asking HEC to help me build the company around it. I need help with business model refinement, pricing strategy, and warm introductions to biotech-focused investors and pharma partners. I am prepared to engage fully with the cohort, to mentor where I have expertise in pharmacology and machine learning, and to be a collaborative member of the community. The venture is pre-seed, pre-incorporation, and pre-revenue. The proof of concept is complete and the results are published on a public benchmark. The next twelve months are about incorporation, the Servier pilot, and the first commercial contracts. HEC Paris Incubateur is the right environment for that next phase. RESEARCH STATEMENT The venture predicts drug resistance mutations from protein sequence alone. No crystal structure is required. The method uses ESM-2 protein language model delta-embeddings combined with ECFP4 drug fingerprints, fed into a Random Forest classifier. The model was validated on the Platinum benchmark, which contains 553 mutations across 53 proteins. Under protein-grouped cross-validation, the model achieves an AUROC of 0.804 with a standard deviation of 0.025. This beats the published state of the art, mCSM-lig, which scores approximately 0.70. On the SKEMPI 2.0 benchmark, the model scores 0.634, which reflects the harder transfer task and guides the next training step. The critical advantage is coverage. Structure-based tools require a resolved crystal structure of the protein-ligand complex. For many clinically relevant targets, no such structure exists. My model covers 100 percent of mutations in the Platinum benchmark, while structure-limited tools cover approximately 18 percent. This is a difference in kind, not a marginal improvement. A pharmaceutical company screening a novel target does not know in advance whether a structure will be available. My model removes that dependency entirely. The scientific roadmap has three phases. First, fine-tune ESM-2 on the SKEMPI 3K mutation set. The current model uses a frozen ESM-2 backbone; fine-tuning on a larger, more diverse mutation set should improve transfer performance. The target is an AUROC of at least 0.70 on the expanded benchmark. Second, validate on prospective clinical resistance data from named partners. I have an active discussion with Servier in Suresnes for a pilot, and research relationships in progress at Paris-Saclay I2BC and Institut Pasteur. Third, package the model as a software service for pharma lead optimization pipelines. The technical foundation is the ESM-2 protein language model, which produces embeddings that capture evolutionary and structural information from sequence alone. Delta-embeddings, the difference between wild-type and mutant embeddings, encode the effect of a mutation on the protein's functional landscape. Combining these with ECFP4 drug fingerprints allows the model to learn mutation-drug interactions directly. The Random Forest classifier is interpretable, fast, and well-suited to the tabular feature representation. The current validation status is honest: proof of concept is complete, the venture is not yet incorporated, and there is no revenue. The Platinum benchmark result is a public, reproducible claim. The SKEMPI 2.0 result of 0.634 identifies the gap that the fine-tuning roadmap addresses. I do not claim product-market fit or validated IP. I claim a validated model with a clear path to commercial deployment. The commercial context is urgent. Antimicrobial resistance is projected to cause ten million deaths per year by 2050. Oncology resistance limits the durability of every targeted therapy. Antiviral resistance undermines pandemic preparedness. Every one of these markets requires resistance prediction during drug development. My model addresses all three with a single sequence-based platform. The next milestone is the Servier pilot. A successful pilot, defined as accurate resistance prediction on a Servier internal target, converts directly into an annual recurring revenue contract. The roadmap from pilot to revenue is the core of the business plan I will develop at HEC Paris Incubateur. ESSAY: FOUNDER PROFILE AND FIT WITH HEC VALUES I am a pharmacist who became a machine learning engineer because I saw the same problem from both sides. In pharmacy school, I learned how drug resistance emerges and how it kills patients. In machine learning, I learned that protein language models can predict mutation effects from sequence alone. The venture is the intersection of those two trainings. The HEC values are ambitious, executor, collaborative, knowledgeable, humble, unique. I will address each. Ambitious: I am targeting a market where the incumbent tool covers 18 percent of mutations and I cover 100 percent. I am building a platform, not a feature. Executor: I built the model, ran the validation, and published the benchmark results as a solo founder. I did not wait for a team or a lab. Collaborative: I have active discussions with Servier, Paris-Saclay I2BC, and Institut Pasteur. I know that science advances through partnerships, and I have already initiated them. Knowledgeable: I hold pharmacy training and machine learning engineering skills. I can read a pharmacology assay and a confusion matrix with equal fluency. Humble: My model scores 0.634 on SKEMPI 2.0. I know exactly where it fails and what the next training step is. I do not overclaim. Unique: A pharmacist who builds protein language models and beats published state of the art as a solo founder is rare. That combination is my edge. What I need from HEC is the business layer. I have not incorporated. I have not set pricing. I have not built a sales motion. The four-month program gives me the structure to do all three, with a cohort of founders who will challenge my assumptions and a network that can open doors to biotech investors. I am not looking for scientific validation. The Platinum benchmark provides that. I am looking for business acceleration, and that is what HEC Paris Incubateur offers. CHECKLIST - [ ] Complete HEC Paris Incubateur short application form - [ ] Complete full application form with venture description - [ ] Attach motivation letter (300-500 words, drafted above) - [ ] Attach research statement (400-600 words, drafted above) - [ ] Attach founder essay on HEC values fit (200-350 words, drafted above) - [ ] Prepare pitch deck for jury selection (10-12 slides) - [ ] Prepare one-page executive summary - [ ] Verify Servier pilot discussion status and obtain named contact - [ ] Verify Paris-Saclay I2BC and Institut Pasteur relationship status - [ ] Confirm Platinum benchmark AUROC 0.804 plus/minus 0.025 is reproducible from public data - [ ] Confirm SKEMPI 2.0 AUROC 0.634 is reproducible from public data - [ ] Prepare financial projection for 12-24 months post-pilot - [ ] Prepare incorporation timeline (France, likely SAS) - [ ] Confirm eligibility for HEC program as pre-incorporation founder - [ ] Submit before rolling deadline; confirm next batch intake date EDITOR NOTES - Eligibility risk: HEC Paris Incubateur may require a registered company or a co-founder team. The profile states the venture is pre-incorporation and sole founder. Verify whether pre-incorporation solo founders are accepted, or whether incorporation must occur before the program start. If required, plan for SAS incorporation in France before the batch begins. - Fact verification needed: The Platinum benchmark AUROC of 0.804 plus/minus 0.025 and SKEMPI 2.0 AUROC of 0.634 must be reproducible from public data. The mCSM-lig comparison at approximately 0.70 must be sourced. The 100 percent versus 18 percent coverage claim must be stated precisely with the benchmark definition of coverage. - Personal detail gap: The essay mentions pharmacy school and machine learning training but does not include the founder's institution names, degree years, or any prior work experience. Insert specific institutions and dates to strengthen the founder profile section. The jury evaluates founder profile heavily, so concrete credentials matter. - Partnership verification: The named partners (Servier, Paris-Saclay I2BC, Institut Pasteur, Sanofi Gentilly) are listed as "named partners" but the status of each relationship is unclear. The letter claims an "active discussion" with Servier and "research relationships in progress" at I2BC and Pasteur. Confirm the actual stage of each conversation before submission. A jury may contact these partners. - Financial projection gap: The roadmap mentions ARR from a Servier pilot but no revenue figures, pricing model, or market size estimate appear in the profile. The HEC jury evaluates business model viability. Prepare a simple financial projection with pilot pricing, expected contract value, and 24-month revenue target before the jury presentation.