9  Guide 6 — The AI Research OS (Platform Vision)

9.1 One platform, five agents, the whole research lifecycle

9.2 What it is

The AI Research OS is the orchestrator that connects the five agents into a single, stateful workflow:

Research Idea
      │
      ▼
[1] Research Discovery Agent      → gaps, ideas, proposal
      │
      ▼
[2] Literature Review Agent       → PRISMA review, evidence table, references
      │
      ▼
[3] Experiment Design & Analysis  → design, notebook, figures, stats report
      │
      ▼
[4] Scientific Writing Agent      → manuscript, journal match, reviewer sim, rebuttal
      │
      ▼
[5] Supervisor / Mentor Agent     → roadmap, weekly progress, defence prep
      │
      ▼
   Journal Submission (Phase 3)

Positioning: “Cursor + ChatGPT + EndNote + Research Assistant + Bioinformatics Analyst — for scientists.”

9.3 Why a platform beats five point tools

Point tools Platform
Each tool re-asks for context One project memory (topic, papers, data, drafts) shared by all agents
Outputs are files on a laptop Outputs are linked objects: a gap → a review section → an experiment → a results paragraph
Sold seat by seat Sold as institutional infrastructure ($50k–500k/year)
Easy to churn Sticky: the project history lives inside

9.4 Platform architecture (Phase 2–3)

┌───────────────────────────────────────────────────────────────┐
│  Web app (Streamlit MVP → FastAPI + React)                    │
├───────────────────────────────────────────────────────────────┤
│  Orchestrator: project state machine (idea → review → ... )   │
│  ├─ Agent 1  ├─ Agent 2  ├─ Agent 3  ├─ Agent 4  ├─ Agent 5   │
├───────────────────────────────────────────────────────────────┤
│  Shared services                                              │
│  • LLM gateway (Claude API; model routing; caching; budgets)  │
│  • Search tools (arXiv, PubMed, Semantic Scholar, Crossref,   │
│    OpenAlex; institutional: Scopus/IEEE/Springer)             │
│  • Document store (papers, PDFs, notes) + vector index (RAG)  │
│  • Data/analysis sandbox (code execution, notebooks)          │
│  • Memory (per project / per student)                         │
│  • Evals & audit logs (every citation → source; every decision)│
├───────────────────────────────────────────────────────────────┤
│  Postgres · object storage · task queue · SSO · billing       │
└───────────────────────────────────────────────────────────────┘

9.5 Institutional requirements (what universities will ask for)

  • Data residency / on-prem or VPC deployment.
  • SSO (SAML/OIDC), role-based access (student, supervisor, admin).
  • Academic-integrity controls: AI-use disclosure, edit logs, plagiarism check hooks.
  • Usage & outcome reports: proposals generated, reviews completed, papers submitted.
  • Training & onboarding (this is where the Codanics content is a competitive advantage).

9.6 Pricing the platform

Tier Annual price Seats Notes
Department $20k–50k up to 200 Cloud, shared tenant
University $50k–150k unlimited SSO, reports, training
University + on-prem $150k–300k unlimited Private deployment, support SLA
Enterprise R&D $200k–500k custom Custom pipelines, integrations

Add-ons: training days ($2k–5k/day), custom domain plugins ($20k–100k), evidence-synthesis services.

9.7 Go-to-market for the platform

  1. Land with one agent (Literature Review) in one department via a free pilot.
  2. Expand to Discovery + Writing across the faculty (individual champions → dean).
  3. Institutionalise with SSO, reports and training → university licence.
  4. Upsell Experiment Agent to labs/core facilities and the Supervisor Agent to the graduate school.
  5. Reference customers in Pakistan/GCC → regional expansion (South Asia, MENA, Africa, SE Asia).

9.8 Moats

  • Workflow + memory (not the model): the model is a commodity; the project graph is not.
  • Domain pipelines (agri, bio, medicine) built with real researchers.
  • Community & training (Codanics YouTube/WhatsApp/course) → distribution nobody else has in the region.
  • Auditability: PRISMA trails, citation provenance, reproducible notebooks — what institutions actually buy.

9.9 Team to build it (first 12 months)

Role Count
Founder / product (research background) 1
AI engineers (agents, tools, evals) 2
Full-stack engineer (web app) 1
Domain scientist / customer success 1
Sales (university & enterprise) 1 (from month 6)

Codanics — Python ka Chilla: Build AI Agents

  • Course: https://codanics.com/courses/python-ka-chilla-build-ai-agents/
  • Website: https://codanics.com · YouTube: @codanics
  • WhatsApp channel: https://whatsapp.com/channel/0029Va7nRDq3QxRzGqaQvS3r
  • Complete free course: https://www.youtube.com/watch?v=UE3o-7KRNXs&list=PLaQmdklCcX40