
RecruitSecure AI — semantic candidate search
Semantic candidate search with embeddings and no external AI API, hybrid search on PostgreSQL + pgvector, and CVs encrypted at rest.
What it is
My own SaaS product for semantic candidate search: you search in natural language and the system ranks by meaning, not by word matches. Embeddings are generated by the server itself with Transformers.js: no calls to an external AI API and no per-query cost.
The problem it solves
Keyword screening misses synonyms and context: “React” is not the same as “frontend with IoT experience”. RecruitSecure understands intent and meaning.
How it’s built
- Hybrid search: cosine similarity on pgvector (HNSW index) combined with full-text BM25 in PostgreSQL; 384-dimension multilingual embeddings.
- CV content encrypted at rest with AES-256-GCM and versioned keys so they can be rotated.
- GDPR built in: data export and deletion with password re-verification, cookie consent and an audit log.
- Stripe billing per plan and per seat (usage-based), with SSO and outbound webhooks for enterprise customers.
- Security and quality: CSP headers, rate limiting on every endpoint, and unit and e2e tests with axe.
My own product end to end; paused since July 2026.
Stack: Next.js 16 · React 19 · TypeScript · Supabase (PostgreSQL + pgvector) · Transformers.js · Stripe.
Other work with a related technical focus:
Mirador — LangGraph home-buying agent
A LangGraph agent with 19 tools and human confirmation before any payment, on a transactional backend and a 3D viewer you can walk through.
Cherry Agents — audited skills and subagents for Claude Code
182 skills and 59 subagents for Claude Code and OpenCode from a single source, with a security scanner, a signed manifest and behavioural evals.
Cherry Appointments — booking over WhatsApp with an AI agent
Multi-tenant booking SaaS: customers book by chatting with an AI agent on WhatsApp or Telegram, while the business runs everything from a single dashboard.