Query all your recruiting data in plain English — across ATS, CRM, HRIS, and data warehouses — without replicating a single record. Federated queries execute live at each source, governed and auditable, through CData Connect AI.
No data movement. No custom integrations. 350+ sources unified behind one natural language interface.
flowchart LR
Q["Ask in plain English"]
P["LangGraph agent pipeline\nIntent → SQL → Execute → Analyze"]
C["CData Connect AI\nfederated query execution at source"]
R["Ranked results\n+ AI explanation"]
S1["👥 ATS\nBullhorn / JobDiva"]
S2["🤝 CRM\nSalesforce / HubSpot"]
S3["🗄️ Warehouse\nSnowflake / Postgres"]
S4["🏢 HRIS / Files\nWorkday / Sheets"]
Q --> P
P --> C & R
C --> S1 & S2 & S3 & S4
style Q fill:#dbeafe,stroke:#3b82f6,color:#1e3a5f,font-weight:bold
style P fill:#dbeafe,stroke:#3b82f6,color:#1e3a5f,font-weight:bold
style C fill:#bfdbfe,stroke:#2563eb,stroke-width:2px,color:#1e3a5f,font-weight:bold
style R fill:#dbeafe,stroke:#3b82f6,color:#1e3a5f,font-weight:bold
style S1 fill:#eff6ff,stroke:#93c5fd,color:#1e3a5f,font-weight:bold
style S2 fill:#eff6ff,stroke:#93c5fd,color:#1e3a5f,font-weight:bold
style S3 fill:#eff6ff,stroke:#93c5fd,color:#1e3a5f,font-weight:bold
style S4 fill:#eff6ff,stroke:#93c5fd,color:#1e3a5f,font-weight:bold
Queries are pushed down to each connected source, so results always reflect live data without any replication or movement. A LangGraph orchestrator coordinates specialist agents — search, analysis, bulk operations, and write workflows each handled independently — before CData Connect AI federates the query across sources with schema translation and access control. The platform uses a REST-first architecture for all deterministic paths (sidebar, profiles, schema discovery), reserving MCP exclusively for the AI agent pipeline to keep rate limit budgets independent and latency low.
- Cross-source federated queries — break data silos without moving data; queries execute live at the source across ATS, CRM, HRIS, warehouses, and spreadsheets simultaneously
- Natural language search — ask "Find senior Java developers in Chicago available under $140K" and get ranked, AI-explained results
- AI candidate matching — relevance scores, skill gap highlights, and plain-English match reasoning from the LLM
- Optimized architecture — REST for all deterministic queries (30–50% faster than MCP on common paths); MCP reserved for NL agent queries only; independent rate limit budgets
- Candidate profiles — placement history, journey timeline, skills, and contact details
- Analytics dashboard — skills in demand, placement trends, recruiter performance metrics
- 350+ data sources via CData Connect AI — Salesforce, Bullhorn, Workday, Snowflake, Supabase, Google Sheets, SharePoint, and more
- Pluggable LLM — supports Groq, Gemini, DeepSeek, Mistral, and OpenAI; Groq and Gemini offer free API tiers
- Encrypted credentials — AES-256-GCM client-side encryption; no backend database required
- Query logs and cost tracking — token usage, query history, and per-session cost monitoring
| Candidate Profile | Analytics Dashboard |
|---|---|
 |
 |
Recommended — scripts handle install and browser launch automatically:
- Windows: double-click
start.bat - Mac / Linux: run
./start.sh
On first run, both scripts install dependencies and open the browser. No manual steps needed.
Or manually:
git clone https://github.com/mohsin-cdata/talent-intelligence-platform.git
cd talent-intelligence-platform
npm install
npm run devOpen http://localhost:3000 and complete the 4-step setup wizard.
No .env file required. All credentials are configured through the in-app setup wizard on first launch:
- CData — your CData Connect AI email, Personal Access Token, and endpoint
- Sources — select which data source connections to query
- LLM — choose a provider and enter your API key (Groq and Gemini are free)
- Review — credentials are encrypted with AES-256-GCM and stored locally in your browser
To get a CData Connect AI account and Personal Access Token: cloud.cdata.com
| Layer | Technology |
|---|---|
| Framework | Next.js 14.1 (App Router) |
| UI | React 18 + Tailwind CSS + Radix UI |
| State | Zustand with localStorage persistence |
| Agent pipeline | LangGraph |
| Data layer | CData Connect AI — REST + MCP |
| Auth | AES-256-GCM (Web Crypto API) |
| LLM | OpenAI-compatible SDK (multi-provider) |
Connects to any of 350+ sources supported by CData Connect AI — including Bullhorn, Salesforce, Workday, Snowflake, PostgreSQL, Supabase, Google Sheets, SharePoint, and more. Schema is auto-discovered on first load; no hardcoded table or catalog names.
- Architecture — full system design: agent pipeline, REST-first data access, federated query flow, 5-layer rate limiter, auth, schema cache, and complete file structure
- Business Use Case — use cases, ROI model, and governance overview
This project includes a CLAUDE.md that Claude Code reads automatically on startup. It covers the architecture constraints, key non-obvious behaviors, modification playbooks (swap data source, add LLM provider, add agent node), and a full recreation guide — so you can extend, debug, or adapt the platform without re-explaining the codebase.
MIT — see LICENSE
Mohsin Turki (mohammedmohsint@cdata.com), CData Software