MOTIVATION LETTER The Linux for Decentralized Science programme in 2026 addresses the exact problem my independent research career has been built around: how a researcher without institutional infrastructure can produce rigorous, reproducible, and verifiable science. I have spent the last two years running computational research projects across addiction neuroscience, protein machine learning, and dynamical systems from a personal Linux VPS, with no university compute cluster and no laboratory affiliation. My preprint record, with three sole-authored manuscripts in review at peer-reviewed journals and two pre-registered replication studies, demonstrates that this model works. What it lacks is a decentralized layer for provenance, data persistence, and verifiable computation. That is what I want to build with this programme. My research pipeline already runs on open-source tools end to end. The CCT model, a tripartite pharmacological framework for reward-memory encoding prevention in addiction, uses Python with scipy ODE solvers and PyMC for Bayesian MCMC calibration. All preprints live on OSF and Zenodo. The hERG cardiotoxicity study, which tested whether bipartite persistent homology predicts cardiotoxicity from protein-ligand interface geometry, was pre-registered and powered before data analysis. It found topological features do not beat a plain descriptor baseline, AUROC 0.8426 versus 0.8782. I reported that negative result directly. The ergofluids project, applying Koopman operator methods with Mori-Zwanzig memory kernels to drug transport in tumor tissue, failed its first real-data validation gate against digitized published figures. I reported that failure as a pre-registered outcome rather than reframing it. These are the practices DeSci exists to incentivize. My technical stack fits the programme's requirements. I run production Linux systems with systemd, Caddy TLS, CI/CD pipelines, and automated backup and disaster recovery. I have built four independent DuckDB-based ingest-to-analyze pipelines across life sciences, technology, and social science domains. I self-host local LLM serving with llama.cpp and on-demand model swapping. I have not yet integrated IPFS for content-addressed storage or smart contracts for provenance, but the architecture of my psyche-twin project, a typed multi-scale knowledge graph with an append-only event log, is designed for exactly that extension. The append-only log is a natural fit for blockchain-anchored timestamps. The self-hosted infrastructure is a natural fit for IPFS pinning. The programme's emphasis on democratizing science matches my position as an independent Nigerian researcher. I have no institutional backing. My endorsements come from Kent Berridge at Michigan, Samuel Gershman at Harvard, Nathaniel Daw at Princeton, and Marcelo Mattar at NYU, secured through the quality of my preprints and my arXiv endorsement, not through affiliation. I am enrolled in the M.Sc. Digital Health programme at Hasso Plattner Institute and University of Potsdam starting Winter Semester 2026/27, which will give me formal academic grounding in digital health infrastructure. The DeSci training this programme offers would let me build the decentralized layer my research pipeline needs and demonstrate a reproducible template other independent researchers in Africa can adopt. RESEARCH STATEMENT I am applying to the Linux for Decentralized Science programme to build a decentralized research infrastructure layer for my existing computational pipelines, anchored by the psyche-twin project and extended to my addiction neuroscience and protein ML work. The core deliverable is a Linux-based system that combines IPFS for content-addressed data storage, blockchain-anchored provenance for pre-registration and analysis logs, and verifiable computation for reproducibility. This directly addresses the programme's selection criteria on decentralized storage, smart contracts, and open access. The psyche-twin project is a typed, multi-scale knowledge-graph architecture for self-modeling. It fuses multiple independent evidence streams, including LLM-derived, assessment-derived, behavioral, and document-derived data, into one append-only event log. Disagreement between streams becomes an explicit graph edge rather than being averaged away. A first-person conversational interface sits on top of the graph, generating responses grounded in the graph's current state. The append-only event log is the critical component for DeSci integration. Every event can receive a content identifier via IPFS, and event batches can be anchored to a blockchain for tamper-evident timestamps. The graph itself can be published as an open dataset with verifiable provenance, allowing other researchers to audit the exact sequence of evidence that led to any generated response. The same infrastructure applies to my CCT model in addiction neuroscience. The model is a tripartite framework with three coupled axes: dopaminergic reward prediction error, NMDAR-dependent long-term potentiation, and affective contrast. It is implemented as an ODE system solved with RK45 and calibrated with Bayesian MCMC using PyMC's DEMetropolisZ sampler, with 14 free parameters and literature-elicited priors from a 1,847-record screen. All five pre-registered hypotheses were confirmed, with posterior super-additivity of 13 to 22 percentage points across model versions. The pre-registration, the prior elicitation protocol, the raw literature screen data, and the posterior samples are all candidates for IPFS storage with blockchain-anchored provenance. A reviewer should be able to verify that the analysis I report is the analysis I pre-registered, down to the hash of the data files. My protein ML work, specifically TOPOLOGIX, also benefits. TOPOLOGIX uses ESM-2 protein language model delta-embeddings plus Morgan/ECFP drug fingerprints and a Random Forest classifier to predict drug-resistance mutations from sequence alone. It achieves AUROC 0.804 plus or minus 0.025 on the Platinum benchmark with 553 mutations, and 0.634 on SKEMPI 2.0. It beats structure-based baselines like mCSM-lig at approximately 0.70 while covering 100 percent of mutations versus approximately 18 percent for structure-limited tools. The training data, model weights, and evaluation scripts can be published as a verifiable computation package, allowing anyone to reproduce the benchmark results without trusting my word. The technical implementation plan has three phases. Phase one, infrastructure setup: configure a Linux server with Docker or Podman containers for IPFS, a blockchain node or light client, and the existing Python research stack. Phase two, pipeline integration: modify the psyche-twin event log to write content identifiers to IPFS on every append, and add blockchain anchoring for event batches. Extend the same pattern to CCT pre-registration files and TOPOLOGIX evaluation outputs. Phase three, documentation and template release: publish a reproducible template repository that other independent researchers, especially in Africa, can fork and adapt to their own pipelines. The template will include systemd service files, Caddy reverse proxy configuration, backup scripts, and a CI/CD pipeline for automated testing. I have the operational experience to execute this plan. I run production Linux systems with automated backup and disaster recovery. I have built four independent DuckDB-based ingest-to-analyze pipelines. I self-host local LLM serving with llama.cpp. I have not yet used IPFS or blockchain tooling in production, but the programme's training component is precisely the gap I need to close. The feasibility of the project is high because the infrastructure is additive to my existing pipelines, not a replacement for them. The impact on democratizing science is concrete. Independent researchers in Nigeria and across Africa often lack institutional repositories, persistent identifiers, and verifiable publication trails. A decentralized layer built on open-source Linux tools removes those barriers. My preprint record, including transparent reporting of negative results in the hERG topology study and the ergofluids validation failure, demonstrates the commitment to research integrity that DeSci requires. This programme would let me turn that commitment into infrastructure others can use. EDITOR NOTES - Chosen research line: psyche-twin as the anchor project, with CCT and TOPOLOGIX as secondary beneficiaries. Rationale: psyche-twin's append-only event log and self-hosted LLM infrastructure map most directly to the programme's decentralized storage and verifiable computation criteria. The CCT model's pre-registration and Bayesian calibration data are strong provenance candidates. TOPOLOGIX's benchmark reproducibility is a strong verifiable computation candidate. The hERG and ergofluids negative results are mentioned in the motivation letter as evidence of research integrity, not as active research lines, which is consistent with their status. - Eligibility risk: the programme page at ngelinux.com does not specify eligibility criteria, geographic restrictions, or whether the programme is a training course, a grant, or a competition. The application assumes an open call for independent researchers. Verify the programme's actual format and eligibility before submitting. If the programme is Nigeria-specific, the applicant's Nigerian nationality is an asset. If it is global, the Africa angle still strengthens the application. - Facts to verify: the M.Sc. Digital Health enrollment at Hasso Plattner Institute is listed as Winter Semester 2026/27, which may be after the programme's timeline. Confirm the start date and whether enrollment status matters for eligibility. The employment history lists a National Product Manager role at Synthcare starting March 2026, which may overlap with the programme timeline. Confirm availability for any residency or intensive training components. - Gaps for the applicant to fill: the programme page mentions blockchain and open-source technology but does not specify a curriculum, mentor list, or expected deliverables. The application should be adapted once those details are available. The applicant should also confirm whether the programme expects a project proposal, a portfolio submission, or a technical challenge as part of the application, and adjust the research statement accordingly. - Tone check: the motivation letter opens with the programme name and the problem, not with "I". The research statement leads with the deliverable and the programme's selection criteria. No banned phrases are used. Every sentence contains a concrete fact, number, or named claim from the profile. The negative results are presented as evidence of rigor, not as failures to be explained away.