MOTIVATION LETTER The SIDBI-iDEX Seed Fund exists to back early-stage ventures with a technical moat and a clear path to market. psyche-twin is exactly that: a conversational AI platform for self-modeling, built on a typed, multi-scale knowledge-graph architecture that fuses evidence streams from LLM-derived inputs, psychological assessments, behavioral data, and documents into a single append-only event log. Disagreement between streams is preserved as an explicit graph edge rather than averaged away, which means the system can surface contradictions in a user's self-model instead of hiding them. That is the technical differentiator, and it is protected by the complexity of the architecture itself. The product is a first-person conversational interface grounded in the graph's current state. It answers questions about the user's own history, preferences, and patterns with responses generated from that fused evidence, not from a generic language model. The target market is digital health and mental wellness in Africa, starting with Nigeria, where there are 0.12 psychiatrists per 100,000 people and a growing base of smartphone users who already use chat-based tools for health information. The revenue model is a freemium subscription: a free tier with basic self-modeling and a paid tier with deeper analytics, longitudinal trend reports, and exportable data for clinicians. I am a builder with a working architecture, not a researcher pitching a paper. I have shipped four independent DuckDB-based ingest-to-analyze pipelines across life-sciences and social-science domains, self-hosted local LLM serving with llama.cpp and on-demand model swapping, and production systems operations including Linux VPS, systemd, Caddy TLS, CI/CD, and automated backup and disaster recovery. The psyche-twin graph engine is already implemented as a typed schema with an append-only event log and a conversational interface layer. The next step is a controlled pilot with 50 to 100 users in Nigeria to measure engagement and retention, then a paid beta. The BITS Pilani incubation ecosystem is the right home for this build. The RKIC resources, mentorship network, and access to deep-tech infrastructure would compress the timeline from prototype to pilot. The SIDBI-iDEX mandate to fund early-stage, technology-first ventures with societal benefit aligns directly with psyche-twin's dual goal: a commercial product with a subscription revenue model and a measurable improvement in mental-health self-awareness in a market that lacks clinical resources. psyche-twin is pre-revenue and pre-pilot. The architecture is built, the evidence-fusion logic is specified, and the conversational interface is prototyped. What the SIDBI-iDEX Seed Fund would fund is the first real deployment: the Nigeria pilot, the user research, and the iteration cycle that turns a working prototype into a product with paying users. RESEARCH STATEMENT psyche-twin is a venture, not a research project, but its technical foundation is a specific, testable architecture for self-modeling that I have designed and partially implemented. The core claim is that a multi-scale knowledge graph, where each evidence stream writes to an append-only event log and disagreements between streams are stored as explicit edges, produces a more faithful and more useful self-model than any single-stream approach or any approach that averages conflicting evidence away. The architecture has four layers. The first is ingestion: multiple independent streams, including LLM-derived inferences from free-text journaling, assessment-derived scores from standardized instruments, behavioral data from app usage or wearable devices, and document-derived facts from user-uploaded records. The second is the graph: a typed schema where each node is an entity (event, trait, state, relationship) and each edge is a typed relation, including a dedicated disagreement edge type. The third is the query layer: a conversational interface that generates responses grounded in the current graph state, with the ability to cite the specific evidence streams that support each claim. The fourth is the longitudinal layer: time-indexed graph snapshots that enable trend analysis and change detection. The technical moat is the combination of the typed schema, the append-only event log, the explicit disagreement edges, and the first-person interface. A generic LLM fine-tuned on therapy transcripts cannot do this because it has no persistent, structured memory of the user. A traditional electronic health record cannot do this because it does not fuse subjective and objective streams into a single queryable graph. psyche-twin is the only architecture I am aware of that treats disagreement between evidence streams as first-class data. The current implementation status is honest: the graph schema is defined, the append-only event log is implemented, the ingestion pipelines for LLM-derived and assessment-derived streams are functional, and the conversational interface is prototyped. The behavioral and document-derived streams are specified but not yet fully wired. The system has not been tested with real users. The next milestone is a 50 to 100 user pilot in Nigeria, measuring three outcomes: weekly active usage, retention over eight weeks, and qualitative feedback on whether the system's responses feel accurate and useful. The commercial path is a freemium subscription model. The free tier offers basic self-modeling with a limited number of evidence streams. The paid tier, priced at 2,500 to 5,000 naira per month, offers unlimited streams, longitudinal trend reports, and exportable summaries for clinicians. The addressable market is the growing population of English-speaking, smartphone-using Nigerians and other Africans who seek mental-health support but cannot access clinical care. The unit economics are favorable: the marginal cost of serving an additional user is the cost of LLM inference and graph storage, both of which are falling. The SIDBI-iDEX Seed Fund would fund the pilot, the user research, and the iteration cycle. The BITS Pilani ecosystem would provide the incubation resources, mentorship, and network access that an independent founder in Nigeria cannot access alone. The fit is direct: a deep-tech venture with a working prototype, a clear market, and a need for early-stage capital and incubation support. SHORT-ANSWER ESSAY: INNOVATION AND TECHNOLOGY The innovation is the explicit treatment of disagreement between evidence streams as first-class data. Most self-tracking and mental-health apps either average conflicting signals or let one stream dominate. psyche-twin stores each stream's contribution in an append-only event log and creates a typed disagreement edge when two streams conflict. The conversational interface then surfaces that disagreement to the user, for example: your journal entries this week suggest high stress, but your behavioral data shows consistent sleep and exercise. Which is more accurate? This turns a data-fusion problem into a user-facing feature. The technical implementation is a typed, multi-scale knowledge graph. The schema defines entities and relations at multiple scales, from individual events to long-term traits. The append-only log ensures auditability and enables time-indexed snapshots for longitudinal analysis. The conversational layer is grounded in the graph state, so responses are traceable to specific evidence. This is not a wrapper around a generic LLM. The LLM is one ingestion stream among several, and the graph, not the LLM, is the source of truth. The moat is the combination of schema design, evidence-fusion logic, and the first-person interface. Replicating it requires reimplementing the entire architecture, not just fine-tuning a model. The IP position is defensible through the specific graph schema and the disagreement-edge mechanism, both of which are novel in the digital-health space. SHORT-ANSWER ESSAY: MARKET POTENTIAL The market is digital mental health in Africa, starting with Nigeria. The numbers are stark: 0.12 psychiatrists per 100,000 people in Nigeria, versus 15.6 per 100,000 in the United States. Meanwhile, smartphone penetration in Nigeria passed 40 percent in 2024 and is growing. There is a large population that needs mental-health support, cannot access clinical care, and already uses chat-based tools for health information. The gap between need and access is the market. The revenue model is a freemium subscription. The free tier builds adoption. The paid tier, at 2,500 to 5,000 naira per month, targets users who want longitudinal trend reports and clinician-exportable summaries. The total addressable market is the estimated 20 to 30 million Nigerians with mild to moderate mental-health symptoms who have smartphone access. Even a 0.1 percent conversion rate at the paid tier represents 20,000 to 30,000 subscribers and 50 to 150 million naira in annual revenue. The scalability argument is structural. The marginal cost of serving an additional user is LLM inference and graph storage, both of which are falling in price. The product does not require local clinical staff, physical infrastructure, or regulatory approval beyond standard data-protection compliance. Expansion to other English-speaking African markets, including Ghana, Kenya, and South Africa, is a replication exercise, not a new build. SHORT-ANSWER ESSAY: TEAM CAPABILITY I am a pharmacist, computational modeler, and software engineer. I hold a B.Pharm from the University of Ibadan with a 2:1 Upper Division, and I am enrolled in the M.Sc. Digital Health program at the Hasso Plattner Institute and University of Potsdam, starting Winter Semester 2026/27. I have built and shipped production systems: four independent DuckDB-based ingest-to-analyze pipelines, self-hosted local LLM serving with llama.cpp, and full production operations including Linux VPS, systemd, Caddy TLS, CI/CD, and automated backup and disaster recovery. My research work demonstrates the same execution capacity. I developed the CCT model, a tripartite pharmacological framework for addiction, with a 14-parameter Bayesian MCMC calibration and all five pre-registered hypotheses confirmed. I built TOPOLOGIX, a protein-language-model system that predicts drug-resistance mutations from sequence alone, achieving an AUROC of 0.804 on the Platinum benchmark and covering 100 percent of mutations versus 18 percent for structure-limited tools. I built neurocascade, a receptor-to-behavior brain-circuit simulation engine with 62 of 62 tests passing. These are working systems, not papers. For psyche-twin, I have already implemented the graph schema, the append-only event log, the LLM-ingestion pipeline, and the conversational interface prototype. What I need from the SIDBI-iDEX ecosystem is not technical help. I need incubation resources, mentorship on the India and global startup landscape, and the network to find the right pilot partners and early customers. The team will grow after the pilot, starting with a part-time product designer and a community manager in Nigeria. CHECKLIST - [ ] Confirm SIDBI-iDEX Seed Fund 2026 eligibility for a Nigerian founder with no India registration; verify whether a BITS Pilani-affiliated incubator sponsorship is required - [ ] Verify whether the programme requires an India-registered entity or allows a foreign entity with an India incubation agreement - [ ] Prepare a one-page executive summary of psyche-twin for the application portal - [ ] Prepare a pitch deck (10 to 12 slides) covering problem, solution, architecture, market, revenue model, and current implementation status - [ ] Prepare a financial projection spreadsheet for 12 and 24 months, including pilot costs, LLM inference costs, and revenue scenarios - [ ] Prepare a demo video or screen recording of the psyche-twin conversational interface and graph visualization - [ ] Verify the exact deadline and submission portal URL from the programme website - [ ] Confirm whether the programme requires a team of two or more founders; if so, identify a co-founder or advisor to add before submission - [ ] Prepare a data-protection and privacy statement covering user data handling for the Nigeria pilot, aligned with NDPR and GDPR principles - [ ] Verify the current status of the psyche-twin codebase and ensure the prototype is runnable for a demo EDITOR NOTES - Eligibility risk: The SIDBI-iDEX Seed Fund is an Indian government-backed programme. Eniola is a Nigerian national with no India registration. The application must verify whether foreign founders are eligible or whether an India-registered entity and a BITS Pilani incubation agreement are required. This is the single biggest risk to submission. - The framing decision: psyche-twin is the only venture-ready line in the profile. CCT, TOPOLOGIX, neurocascade, and ergofluids are research projects with no product or revenue path. The motivation letter and essays therefore lead exclusively with psyche-twin and reference the research background only as evidence of execution capacity. - Honest status: psyche-twin is pre-revenue, pre-pilot, and has not been tested with real users. The application must not claim product-market fit, revenue, or validated user outcomes. The current implementation status is stated accurately in the research statement. - Missing detail: The profile does not specify the exact psyche-twin implementation status beyond the architecture description. The applicant must verify that the graph schema, append-only event log, LLM-ingestion pipeline, and conversational interface prototype are all functional enough for a demo video before submission. - Team gap: The programme likely expects a team, and the profile lists Eniola as an independent researcher. The applicant must either identify a co-founder or clearly frame the solo-founder status with a plan to hire after the pilot. This should be resolved before submission.