CHAPTER 1
AI KNOWLEDGE SYSTEM
How I turned 9 years of oncology expertise into a living, breathing intelligence system.
The Problem
For years, the organization carried its knowledge in fragments:
PDFs on desktops, WhatsApp notes from clinicians, PowerPoints buried in inboxes, and medical research scattered across personal drives.
Every new employee — clinician, technician, coordinator — had to swim through oceans of scattered information to understand:
How DNA Ploidy works
How LABS are set up
What Hospitals Need
How sample collection is done
What doctors ask
How to respond to patients
How to pitch to cancer institutes
How cases should be escalated
Tribal knowledge was drowning in chaos.
And scaling cancer prevention requires clarity, not confusion.
The future doctors, partners, and employees needed a single source of truth.
My Mission
Build an AI system that doesn’t just answer questions —
but trains people, aligns teams, and compresses a decade of learning into seconds.
A system that:
Understands the science
Understands the product
Understands the operations
Understands the business
Speaks like the company itself
This was not an “AI assistant.”
This was institutional memory with a heartbeat.
My Role & Responsibilities
I designed the architecture of the entire knowledge universe:
Strategy 90%
What the AI should know
How it should interpret questions
How it should respond to clinicians vs front desk staff
How technical its language should be
How it should behave under uncertainty
What datasets it should prioritize
How to prevent misinformation
How to ensure 100% factual accuracy
Content & Research 80%
I manually reviewed hundreds of documents including:
Lab SOPs
Clinical guidelines
DNA Ploidy research
Cervical & oral cancer workflows
Hospital pitch decks
FAQs from doctors
Educational decks
Operational sheets
Billing processes
Medical equipment instructions
On-ground staff learnings
This wasn’t prompting — this was knowledge engineering.
I converted PDFs to editable formats, extracted relevant content, validated all medical terminology with available references, and built a structured taxonomy of:
Concepts
Procedures
Risks
Exceptions
Definitions
Process flows
Quick responses
Sensitive-answer rules
This became the brainstem of the AI.
Design 100%
I created the information architecture:
Tagging System
Topic clusters
Medical hierarchies
Operational workflows
Prompt scaffolding
Domain-context routing
Guardrail frameworks
This ensured the AI could:
Differentiate between a doctor’s question vs a receptionist’s
Provide operational answers vs clinical summaries
Maintain compliance tone
Avoid medical diagnosis claims
Provide clarity even when users ask unclear questions
I designed it like a knowledge product, not a chatbot.
Execution 75%
I led the execution from raw data → polished intelligence:
Broke down long PDFs into structured chunks
Cleaned formatting
Removed redundant or outdated content
Created version control for every knowledge update
Designed accuracy-testing frameworks
Built a looping system for error detection
Logged inconsistent answers
Retrained and refined until reliability hit near 100%
The system was tested under:
Medical-use scenarios
Operational scenarios
Sales scenarios
Onboarding scenarios
Emergency troubleshooting scenarios
This allowed the final AI to behave like a cross-functional expert.
The Result
The AI system became:
The company’s first internal training engine.
New employees reduced their learning curve from months to days.
A source of truth for field teams.
Hospital staff could quickly understand procedural workflows.
A rapid support tool for future doctors.
No more searching 40-slide decks for one answer.
No more inconsistent guidance.
A business enabler.
Sales teams used it for pitch clarity.
Clinicians used it to cross-verify facts.
Lab teams used it to follow SOP standards.
Long-term infrastructure.
Every new lesson, every new lab, every new doctor can now be added to the system — ensuring knowledge never disappears, even if people do.
The company now owns its intelligence.
