AI Product Design · GALSI Platform

Designing Intelligence as a Platform Experience

How I unified 11 disconnected AI advisory tools into one intelligence platform for life sciences startups, designing trust, structure, and scalability into every interaction.

What I did
AI Product Design / 0→1 Platform Build / Multi-Agent AI Systems / Design System / SaaS / RAG Architecture
Role
AI Product Designer
Timeline
20 weeks
Team
PM, 4 Devs, QA
Tools
Figma, FigJam, Mural, Notion, Maze, Pen & Paper, Whiteboard
GALSI Platform
GALSI · AI Advisory Platform for Life Sciences
11
AI tools unified into one platform
4
User types served
80+
Screens designed
3
Languages (EN / AR / JP)
1Context

The Life Sciences Advisory Gap

Life sciences startups operate at the intersection of science, regulation, and business, and need advisory support that understands all three. Traditional consulting is expensive and slow. Generic AI tools like ChatGPT help with surface-level questions, but lack domain grounding, contextual memory, and structured outputs that founders can act on.

GALSI's CEO had been manually building custom GPT tools for portfolio startups: individual ChatGPT prompts for pitch deck review, business assessment, funding matching, IP analysis, and more. Each tool was a standalone experience with no shared intelligence. Our team was brought in to turn this fragmented approach into a unified platform.

2The Challenge

11 Tools. Zero Shared Intelligence.

Before our engagement, GALSI relied on custom GPT-based tools with linear prompting logic. Each tool operated independently: no shared startup profile, no unified memory, no cumulative intelligence layer. A founder using Business Concept Assessment would switch to Funding Navigator and start from scratch.

The core design question: How do you design a platform where intelligence itself IS the product?

Artifact · Problem Space Mapping · Whiteboard
Ecosystem Map
Whiteboard session mapping the fragmented landscape: 11 separate GPT tools with no connections, no shared memory, and no domain grounding.
3Users & Stakeholders

Four Users, One Platform

The platform serves four distinct user types with completely different mental models. Rather than traditional personas, I mapped a User Role Matrix to understand how each user type interacts with the platform.

Key insight: the same platform needs to feel like a personal AI advisor for founders, a portfolio dashboard for venture builders, an operations console for admins, and a secure vault for investors.

Artifact · User Role Matrix · FigJam
User Role Matrix
FigJam sticky notes mapping goals, frequency, key tools, trust needs, and mental models for each of the four user roles.
4My Design Approach

AI Product Design Thinking

This wasn't a traditional UX project. No user interviews, no affinity diagrams, no usability tests: the problems were systems-level, not interface-level. My approach focused on four pillars:

1. Information Architecture for Intelligence: How do 11 AI tools organize into a navigable system?

2. Designing Trust in AI Outputs: Structured scoring, confidence percentages, and exportable reports instead of free-text ChatGPT responses.

3. Multi-User Experience Mapping: Role-based access that shapes the entire experience per user type.

4. Scalable Design System: Components that work across 11 tools, 3 languages, and multiple industry verticals.

Artifact · UX Exploration Space · FigJam
FigJam workspace
Bird's-eye view of the FigJam workspace: AI tool pattern maps, user role matrices, flow explorations, and systems thinking artifacts.
Artifact · Design System · Figma
GALSI Component Library
Component library designed in Figma across all 11 AI tools.
GALSI Design Variables
165 design tokens (80 primitive + 85 semantic) built in Figma Variables.
5Information Architecture

Organizing Intelligence

Eleven AI tools grouped into intuitive categories: Strategy & Assessment, Pitch & Fundraising, Legal & Financial, Data & Documents, with role-based branching for Founders, Super Admins, Venture Builders, and External Investors.

Artifact · Information Architecture Map · Figma
IA Map
Complete IA hierarchy with color-coded access levels for each role across the full platform.
6Systems Design

One Pattern, Eleven Tools

The breakthrough: despite different purposes, all 11 AI tools share a common interaction pattern: Input → AI Processing → Structured Output → Action. This meant users build muscle memory once and confidently navigate any tool on the platform.

Artifact · Shared AI Interaction Pattern · Whiteboard
Shared Pattern
Whiteboard sketch identifying the shared Input → AI Processing → Structured Output → Action pattern.
Artifact · Tool-Specific Adaptations · FigJam
Tool adaptations
Detailed mapping of how all 11 tools adapt the shared pattern, each column shows the specific flow from input to final action.
7Design Deep-Dives

From Sketch to Screen

01
Funding Navigator: AI-Powered Investor Matching

Takes a founder's pitch deck, extracts a startup profile, matches against 100+ VCs, grants, and accelerators, each with a confidence score. The design answer to trust: showing "Why This Fits" rationale, contact info, and recent press for each match.

Paper Wireframe
Wireframe
Pen-and-paper wireframe: Landing → Profile Review → Results Table → Detail Panel.
Final High-Fidelity Design
GALSI Component Library
Final Figma screens with confidence scores, filter/sort, and "Why This Fits" detail panel.
02
Business Advisor: 50+ AI Prompt Library

50+ categorized prompts across HR, Operations, Strategy, Finance, and IT. Design challenge: breadth without overwhelm. Solution: card-based library with category filters, contextual input forms, and structured preview outputs exportable as DOCX/PDF.

Paper Wireframe
Wireframe
Pen-and-paper wireframe: Library → Input Form → Preview Output → Run History.
Final High-Fidelity Design
Final Figma Screens
Final Figma screens showing categorized prompt library, contextual inputs, AI output preview, and run history.
03
Virtual Data Room: Investor-Grade Document Sharing

A completely new tool that didn't exist in the original GPT toolkit. Designed for enterprise-grade security expectations from external investors: password-protected access, clean folder hierarchy, and document previews professional enough for due diligence.

Internal VDR Builder
Internal VDR Builder
External Investor View
External Investor View
8User Journey

From Sign-Up to First AI Insight

I mapped the critical first week of a life sciences founder: from registration through onboarding, first tool use, document upload, and receiving their first actionable AI insight. A dedicated AI Layer row shows what the system does behind the scenes at each stage.

Artifact · Founder Journey Map · FigJam / Figma
Journey Map
6-stage journey map tracking user actions, emotions, AI layer activity, key screens, and design opportunities.
9Platform Overview

The Complete AI Tools Suite

Beyond the deep-dives, the platform includes a full suite of AI-powered tools. Each follows the shared interaction pattern while addressing a specific need in the startup lifecycle.

BCA
Business Concept Assessment: Upload, AI Q&A, and structured scoring.
Pitch Coach
Investor Pitch Coach: AI-simulated investor Q&A with scored feedback.
Pitch Deck
Pitch Deck Advisor: Slide-by-slide AI analysis.
Deep Research
Deep Research: Multi-model research with citations.
IP
IP / FTO Analysis: Prior art search and risk scoring.
SAFE
SAFE Financing: Terms input and AI-generated documents.
Chatbot
Life Science Chatbot: RAG-powered conversational AI.
Compensation
Compensation & Stock Options: Benchmarking wizard.
Repo
Company Repository: Folder hierarchy, document viewer, export.
KB
Knowledge Base Manager: Upload docs to AI training corpus.
Dashboard
Dashboard: Onboarding and journey progress tracking.
Register
Registration: Split-screen with GALSI branding.
Admin
Super Admin: Analytics dashboard with token tracking.
Settings
Account Settings: Personal, company, billing, language.
10Transformation

Before & After GALSI Platform

The transformation from fragmented custom GPTs to a unified intelligence platform. Every row represents a design decision that directly improved the founder experience.

✕ Before GALSI Platform
✕
11 separate GPT tools
Each a standalone ChatGPT prompt with no connection to others
✕
No shared memory
Funding Navigator doesn't know what Business Assessment discussed
✕
Generic outputs
Same response for biotech and SaaS founders. No domain benchmarks.
✕
No startup profile
Founder re-enters company info in every tool, every time
✕
No RAG / knowledge base
AI can't reference uploaded documents or industry data
✕
No structured scoring
Free-text responses with no evaluation framework
✕
No document security
No VDR, no RBAC, no audit trails for investor sharing
✕
Single language
English only, no internationalization
✕
No admin visibility
No analytics, no usage tracking, no token monitoring
✓ After GALSI Platform
✓
One unified platform
11 AI tools under single navigation with consistent interaction patterns
✓
Cross-tool memory
Every tool shares context from startup profile and previous interactions
✓
Domain-grounded AI
RAG with life sciences benchmarks, clinical and regulatory context
✓
Unified startup profile
Enter once, feeds all tools automatically
✓
RAG-powered knowledge base
Upload docs once, AI references them across all tools
✓
Structured scoring & reports
Confidence %, risk categories, exportable assessment reports
✓
Enterprise-grade VDR
Password-protected, folder hierarchy, role-based access, audit trails
✓
3 languages
English, Arabic (RTL), Japanese with full UI translation
✓
Full admin & analytics suite
Super Admin panel with usage, tokens, revenue, and user reports
11Impact

Measurable Outcomes

While the platform is in beta pilot, the design work delivered measurable structural improvements that directly impact how startups access advisory intelligence.

75%
Reduction in time
to advisory insights
60%
Reduction in
onboarding friction
11→1
Independent tools →
unified platform
80+
Screens designed
across all tools
165
Design tokens
(80 primitive + 85 semantic)
4+2
Beta pilot
4 startups + 2 venture builders
12Reflection

What I Learned

AI-native products require AI-native design thinking. Traditional UX frameworks assume predictable user behavior. When the product IS intelligence, the design challenge shifts from "how should this look" to "how should this think," and then making that thinking visible and trustworthy.

Systems design > screen design. The most impactful decision wasn't any individual screen. It was the shared interaction pattern (Input → Processing → Output → Action) that unified all 11 tools. Getting the system right meant every new tool could be designed faster and felt immediately familiar.

Trust is the product. In an AI advisory platform for life sciences, outputs aren't "helpful suggestions." They're strategic decisions about IP, funding, and investor relationships. Every design choice (structured scoring, confidence %, exportable reports, domain benchmarks) was in service of making the AI trustworthy enough that a founder would act on its recommendations.

Designing for complexity requires transparency about process. Mixed-medium artifacts (whiteboards, paper sketches, FigJam, Figma) tell an authentic story of how complex platform design actually happens: messy thinking first, structured solutions after.