How AI Powers It

• RAG pipeline: Documents are chunked, embedded, and stored in a local vector database (e.g., Qdrant or ChromaDB)
• Local LLM inference: Uses fast quantized models (leveraging BaseRT/llama.cpp-style local inference) so it runs without internet on government hardware
• Hybrid search: Combines semantic vector search with keyword search to handle military jargon and acronym-heavy queries accurately
• Citation grounding: Every answer is grounded in retrieved source chunks, dramatically reducing hallucination risk — critical for operational decisions

Target Customers (MVP Go-to-Market)

• Defense contractors and consulting firms supporting DoD training commands
• Professional Military Education (PME) institutions: Command & General Staff College, War College, ROTC programs
• National Guard units (often underserved by large prime contractors)
• Individual units purchasing through GSA Schedule or OTA agreements

Unique Selling Points

• Zero cloud dependency: Runs entirely on-prem or air-gapped — the #1 requirement that disqualifies every general-purpose AI tool
• Pre-loaded military knowledge: Not a generic RAG tool; ships ready-to-use with the doctrine corpus soldiers actually need
• Unit memory layer: Each unit builds its own private knowledge layer on top of shared doctrine — the system gets smarter the longer a unit uses it, surviving personnel turnover
• Citation-first answers: Every response is traceable to a specific FM, AR, or SOP — meets the military's accountability and verification culture
• Small team buildable: Entire MVP is a Python RAG stack + React UI + Docker packaging — two engineers, 10 weeks

Revenue Model

• Perpetual license + annual support: $8K–$25K per unit/organization depending on size (fits within micro-purchase and simplified acquisition thresholds)
• Contractor seat licenses: $200/seat/month for defense consulting firms
• PME institutional license: $15K/year for training schools with unlimited users

Why This Wins

• No classified data required for MVP — eliminates the biggest procurement barrier
• Public doctrine corpus is freely available — zero data licensing cost
• Faster sales cycle targeting contractors and PME schools vs. direct DoD procurement
• Strong defensible moat: the pre-built military doctrine corpus + unit memory layer creates compounding switching costs the longer a unit uses it

How AI Powers It

• RAG (Retrieval-Augmented Generation) pipeline using LlamaIndex or LangChain over a vector database (Pinecone or Qdrant) — ensures answers are always grounded in uploaded documents
• Strict citation mode enforced at the prompt level — the model is instructed to refuse to answer if supporting evidence cannot be found in the document corpus, making it safe for engineering use
• Semantic conflict detection — embeddings are used to identify semantically similar procedures across documents, then a reasoning layer checks for contradictory values or supersession language
• Revision-aware indexing — document metadata (revision date, document number) is extracted on upload so the system can reason about which version is authoritative

Tech Stack for MVP (Small Team Friendly)

• Next.js frontend (simple chat UI + document manager)
• Python FastAPI backend
• LlamaIndex for RAG orchestration
• Qdrant (open source) for vector storage
• GPT-4o or Claude 3.5 Sonnet as the reasoning layer
• Supabase for auth, metadata, and audit trail storage
• PDF parsing via PyMuPDF + unstructured.io

Total build time for MVP: 6–8 weeks for a team of 2.

Unique Selling Points

• Aerospace-specific, not generic — purpose-built for the document types, terminology, and compliance context of aerospace; not a horizontal doc tool bolted onto the industry
• Citation-only safety guarantee — markets itself explicitly as "no hallucinations, only citations," which is the #1 concern for safety-critical engineering teams
• Conflict detection is a unique moat — no other lightweight tool automatically flags AD/SB conflicts with AMM procedures
• Exportable audit trail — turns AI queries into compliance-ready work records, directly reducing audit prep time
• ITAR-safe option — roadmap includes a single-tenant or on-prem deployment for customers with export-controlled documentation

Go-To-Market

• Target customer: Quality managers and lead engineers at FAA Part 145 repair stations (5,000+ in the US alone) and AS9100-certified aerospace suppliers with 10–200 employees
• Pricing: $199/month per workspace (up to 10 users, 5GB document storage) — affordable compared to legacy MRO software at $50K+/year
• Acquisition: Direct outreach to Part 145 repair station directories, aerospace LinkedIn communities, and MRO conference presence (MRO Americas)
• Land and expand: Start with the quality/compliance team, expand to engineering and production floor technicians

Potential Impact

AeroLens compresses hours of document research into seconds, reduces compliance audit preparation time by an estimated 60%, and creates a living institutional knowledge base that survives engineer turnover — directly improving safety outcomes and operational efficiency in one of the world's most documentation-intensive industries.