Services
How I Can Help
My work covers the whole product: the data model, the API that decides who can see what, the background jobs that bill and notify, and the interface on top. AI features sit behind one abstraction, so a voice agent can move between Retell and x.AI Grok realtime with zero downtime, and billing runs are idempotent, so a retry never double-charges.
AI Voice Agents
Inbound call automation with real-time tool calling for bookings, and transcripts saved to the CRM. Retell, x.AI Grok realtime, Twilio Media Streams and WebSockets.
LLM Features in Production
Tool calling, retrieval-augmented generation and vector search inside real products, with validated, logged results. pgvector, embeddings and prompt design.
Multi-Tenant SaaS Products
Platforms where the backend is the single authorization boundary for every role. FastAPI, NestJS, Rails, PostgreSQL, Auth0 and RBAC.
Billing & Payments
Idempotent billing engines and subscription plans that never double-charge, on Stripe and background jobs (Celery, pg-boss).
Real-Time Platforms
Live leaderboards, scoring and dashboards that stay in sync across processes, with Socket.IO, WebSockets and PostgreSQL LISTEN/NOTIFY.
Ruby on Rails Development
New Rails products and feature work on existing Rails codebases, with access rules, RSpec coverage and reviewed releases. Devise, Sidekiq, RuboCop and Haml.
Legacy Rewrites & Hardening
Audit, fix root causes, add tests that prove it, and ship through a CI-gated pipeline with pytest, Vitest, Playwright and GitHub Actions.
What working together looks like
The backend is the boundary
Every request resolves its tenant and role on the server before a scoped query runs, so a tenant only ever sees its own data.
Providers behind one switch
Voice and LLM providers sit behind one abstraction, so changing vendors doesn't mean a rewrite or downtime.
Money paths are idempotent
Billing runs carry an idempotency key, so a retried run produces one invoice, never two.
Root cause, tests, gated deploy
Audit, fix the root cause, add tests that prove it, and ship through a CI-gated pipeline.
Building a SaaS product with AI in it?
From the data model to the deployed interface: multi-tenant platforms, AI voice agents, LLM features and Stripe billing.