BA/RE/PM · consulting.bridging-data.com · Serverless

AI Consulting Accelerator

Workflow: Discovery → Analysis → Delivery
Skills: 17 named, versioned skills
Knowledge: 26 industry frameworks (PDF/HTML)
Status: Live · private demo
System Architecture
Clientprivate demo
consulting.bridging-data.comNext.js SPA · single-admin Cognito login
·
Dashboard · Framework Q&A · Discovery · Analysis · Delivery · Engagements
CDN & DNS
Route 53*.bridging-data.com
→
CloudFrontACM wildcard cert
→
S3 StaticNext.js static export
API GatewayCognito JWT auth
JWT AuthorizerCognito single-admin pool
→
OPTIONS preflightno auth (CORS)
→
FastAPI Lambdaall protected routes
LambdaFastAPI + Mangum
consulting_apiroutes: query · structure · run · stakeholders · engagements · skills · sources
·
Skill Registry17 named skills — invoked by name, never by similarity
Core Libraryaiplatform/ (shared)
PDF/HTML Loaders26 frameworks
→
Chunker800 tok / 100 overlap
→
SHA-256 Dedup
→
OpenAI Embedder
pgvector SearchHNSW · cosine
→
Language LockDE/EN server-side detect + filter
→
LLM → Cited Draft
Knowledge Storepgvector · scoped
consulting26 framework PDFs/HTML
skills66 SKILL.md files (org memory)
projects7 project READMEs (org memory)
radar · competitor · regulatoryweekly reports from S3 (read-only)
Engagement Storeisolated DB
CONSULTING_ENGAGEMENT_DB_URLconfidential data — never shared with public db-v2
·
Engagement ORMturns (JSONB) · extras (JSONB) · assessment · archived
External
OpenAIgpt-4o-mini · text-embedding-3-small
Anthropic Claudeswappable LLM provider
Amazon Cognitosingle-admin pool · no self-signup
S3 (read-only)radar/competitor/regulatory reports → org memory
Discovery → Analysis → Delivery — 17 Named Skills
Discovery 8 skills
Business Problem Structurer
consulting.structure-business-problem v1.4
Situation, gap, pain points, goals/metrics, root cause, stakeholders, scope
Stakeholder Analysis
consulting.analyze-stakeholders v1.2
RACI skeleton, influence/interest grid, engagement levels, comms plan
Risk Identification
consulting.identify-risks v1.0
Risk · Impact · Probability · Recommendation
Assumption Detection
consulting.detect-assumptions v1.0
Implicit assumptions to surface and validate
Open Questions
consulting.open-questions v1.1
Context-aware clarification questions referencing the user's input
Hypothesis Generation
consulting.generate-hypotheses v1.1
Evidence-grounded, confidence-ranked root-cause hypotheses
Interview Guide
consulting.interview-guide v1.0
Stakeholder discovery interview structure
AI Readiness Assessment
consulting.assess-ai-readiness v1.0
Data · tech · talent · process · governance → blockers + first steps
Analysis 7 skills
Requirements Structurer
consulting.structure-requirements v1.4
Goal / requirement / user-story split · INVEST stories · quality flags
Refine Analysis (Delta)
consulting.refine-analysis v1.0
Delta-aware Updated Findings from new answers — used in engagement rounds
Consultant Assessment
consulting.consultant-assessment v1.0
Preliminary assessment + priorities (risks/impact/quick wins)
Pattern Matching
consulting.match-patterns v1.0
0–2 resembling project archetypes from curated catalog — hypotheses to validate, never prescriptions
Relevant Knowledge
consulting.relevant-knowledge v1.0
Cross-source internal references (skills, projects, radar) — cited, never invented
Compliance Maturity
compliance.assess-maturity v1.0
NIST AI RMF / GDPR-DSG / AWS WA Security → prioritised gap analysis
AI Governance
consulting.assess-ai-governance v1.0
EU AI Act risk class + NIST AI RMF governance gaps — org structures, not tech controls
Delivery 2 skills
Roadmap + Backlog
consulting.structure-roadmap v1.0
Now/Next/Later roadmap · Epic → Feature → User Story agile backlog (Scrum/roadmapping grounded)
Value Articulation
consulting.articulate-value v1.0
Technical initiative → executive narrative: transformation story, KPI mapping, stakeholder benefits — no invented figures
Framework Q&A RAG
Framework Q&A
POST /api/v1/consulting/query
IREB · BABOK · BPMN · PMBOK · Scrum · EU AI Act · GDPR · NIST · AWS WA — grounded answers with source citations
Organizational Memory — Cross-Source Knowledge Retrieval
skills (66 SKILL.md) cap: 4
projects (7 READMEs) cap: 1
radar (weekly reports) cap: 1
competitor (weekly) cap: 1
regulatory (monthly) cap: 1
→
derive_search_keywords()
business context → technical keyword bag
retrieve_knowledge()
embed once · multi-app_name query · threshold 0.25
→
Cited references to validate skill:// · project:// · radar:// · competitor:// · regulatory://
Honest "none found" threshold-gated — no fabrication, no hallucination
Vocabulary bridge capability metadata (description: frontmatter) + query-side keyword expansion — bridges business ↔ technical language gap
Engagement Flow — Stateful Multi-Round Discovery
Technology Stack
LayerTechnology
Backend API FastAPI Mangum Python 3.12 uv
Frontend Next.js TypeScript Tailwind v4 lucide-react
Vector Storage PostgreSQL 16 pgvector HNSW index AWS RDS (shared db-v2)
Engagement DB PostgreSQL (isolated) SQLAlchemy async JSONB turns/extras
Embeddings OpenAI text-embedding-3-small 1536-dim
LLM (swappable) OpenAI gpt-4o-mini Anthropic Claude
Document Processing pdfplumber BeautifulSoup4 python-docx fpdf2
Shared Library aiplatform — loaders, chunker, embedder, vector store, LLM providers, settings
AWS Infra Lambda (container) API Gateway Cognito S3 CloudFront Route 53 ACM
Quality eval.py — 24 golden cases · rule-based checks · LLM-as-judge (gpt-4o) · 78 tests green