Inside the Black Box
You deserve to know exactly how the AI that manages your health actually works — what it knows, where it learned it, how it checks itself, and how it gets smarter the longer you use it.
What Powers SciAgent
Not a chatbot. A clinical reasoning engine trained on the science of inflammation.
SciAgent is built on a large language model — the same class of AI behind the most advanced reasoning systems in the world — but fine-tuned specifically for oral-systemic health. Think of the base model as a medical school graduate: broad knowledge, strong reasoning. Our fine-tuning is the residency — years of specialized training in the exact pathways that connect your mouth, your gut, and your systemic inflammation.
Foundation Model
A state-of-the-art language model provides the reasoning backbone — the ability to read your data, recognize patterns, weigh tradeoffs, and explain its thinking in plain language. We don't build this from scratch. We build on the best.
Domain Fine-Tuning
We then train the model on oral-systemic health specifically: periodontal pathology, gut barrier dynamics, inflammatory cascades, biomarker interpretation, pharmacology of our product lines. It doesn't guess about inflammation. It reasons from mechanism.
Constrained Outputs
SciAgent can only recommend interventions from the Fasolati product ecosystem. It cannot prescribe medications, diagnose diseases, or replace your doctor. It operates within strict clinical guardrails — what to rinse, swallow, and avoid. Not medical advice. Protocol guidance.
Explainable Reasoning
Every recommendation SciAgent makes comes with a "why." Not a black box output — a transparent chain of reasoning you can read, question, and share with your clinician. If it suggests increasing your Loria frequency, it tells you exactly which data point triggered that change.
Where It Learned What It Knows
Trained on peer-reviewed science. Not internet health forums. Not sponsored content.
The quality of an AI is only as good as the quality of its training data. SciAgent's knowledge base is built from curated, peer-reviewed clinical literature — the same sources your periodontist and gastroenterologist rely on. We don't scrape the internet. We curate the science.
Clinical Literature
Thousands of peer-reviewed papers on periodontitis, gut permeability, inflammatory biomarkers, microbiome ecology, and the oral-systemic connection. Sources include The Lancet, Nature, Journal of Clinical Periodontology, Gut, and specialized inflammation research.
Biomarker Databases
Reference ranges, population distributions, and clinical significance thresholds for hs-CRP, HbA1c, zonulin, calprotectin, LPS-binding protein, and other inflammatory markers. SciAgent knows what your numbers mean in context — not just whether they're "high" or "low."
Product Pharmacology
Deep knowledge of every Fasolati formulation — mechanisms, dosing, interactions, contraindications, expected response curves. When SciAgent adjusts your Lytica protocol, it understands the enzymatic and phage mechanisms at the molecular level.
Clinical Practice Guidelines
Current treatment guidelines from the American Academy of Periodontology, American Gastroenterological Association, and the European Federation of Periodontology. SciAgent's recommendations align with — never contradict — established standards of care.
What SciAgent is never trained on: social media health claims, supplement marketing, anecdotal testimonials, or unreviewed preprints. If it's not in the peer-reviewed literature or validated clinical databases, SciAgent doesn't know it — and that's by design.
How It Verifies Everything
Every recommendation runs through three layers of verification before it reaches you.
We built SciAgent with a fundamental assumption: AI makes mistakes. The question isn't whether errors happen — it's whether the system catches them before they reach you. Our answer is a three-layer verification architecture that treats every output as unproven until confirmed.
Layer 1
Evidence Match
Every claim is cross-referenced against its source literature. If SciAgent says "hs-CRP above 3.0 correlates with increased cardiovascular risk," it must cite the specific studies.
Layer 2
Clinical Guardrails
Rule-based safety checks catch anything outside established parameters. Dosing limits, contraindication flags, interaction warnings — hard-coded, not learned. The AI cannot override these.
Layer 3
Confidence Scoring
Every recommendation carries a confidence score. High confidence: act on it. Medium: SciAgent explains the uncertainty. Low: it flags the question for your clinician instead of guessing.
Never Invents
SciAgent cannot fabricate studies or statistics. It retrieves from its verified knowledge base or says "I don't know."
Never Overreaches
If your data suggests something outside its scope — a possible diagnosis, a medication question — it routes you to your doctor, not a guess.
Always Auditable
Every recommendation is logged with its reasoning chain. Your clinician can review exactly why SciAgent made each suggestion.
How It Learns You
The longer you use Fasolati, the smarter your personal model becomes.
SciAgent starts with population-level science — what works for most people. But you're not most people. Over weeks and months, it builds a model of your specific inflammatory patterns, your response curves, your habits, and your triggers. This is where it transforms from a general tool into your personal health engine.
Week 1 — Baseline
SciAgent ingests your initial data: lab results, oral exam findings, any wearable data you connect. It establishes your personal baseline across the three vectors — Oral Ecology, Gut Integrity, Systemic Inflammation — and sets starting protocols based on population evidence.
Weeks 2–4 — Response Mapping
As you use the products and log data, SciAgent begins tracking how your body responds. How quickly does your hs-CRP drop after a Lytica intervention? Does your sleep quality correlate with gut protocol adherence? It's building your personal response curves.
Months 2–3 — Pattern Recognition
With enough data, SciAgent starts identifying your unique patterns. Maybe your inflammation spikes every time you travel. Maybe your oral ecology dips on weekends. Maybe Biolumen works faster for you than population averages predict. These insights shape increasingly precise protocols.
Month 4+ — Predictive Protocols
This is where the model matures. SciAgent begins anticipating inflammatory events before they show up in your biomarkers — adjusting protocols proactively based on behavioral patterns, seasonal trends, and your established response curves. It moves from reactive to predictive.
Ongoing — Continuous Refinement
Your model never stops learning. Every new lab result, every oral exam, every week of wearable data sharpens the picture. And as the broader Fasolati user base grows, anonymized population insights feed back into the system — your model benefits from the collective without ever exposing your data.
What Happens to Your Information
Your health data is yours. Period.
Encrypted at Rest
Your personal health data is encrypted with AES-256 encryption — the same standard used by banks and defense systems. Even our engineers cannot read your individual records.
Never Sold
Your data is never sold, rented, or shared with advertisers, insurance companies, or employers. Fasolati's business model is products and subscriptions — not your information.
You Control It
Export your complete health record anytime. Delete your data anytime. Pause data collection anytime. Your model, your data, your decision.
De-identified Research
If you opt in, anonymized and de-identified patterns from your data contribute to population-level research that improves the system for everyone. No identifying information ever leaves your account.