As a Full Stack Engineering Intern at DMFlo, I’m working across the stack on a production SaaS that automates Instagram DMs, comment-to-DM, and story-reply campaigns for creators and small businesses.
On the infrastructure side, I architected a webhook-driven ingestion pipeline — Meta HMAC-SHA256 verification → account resolution → BullMQ enqueue — that decouples inbound Instagram events from downstream processing. Load-tested to 90+ req/s at 0% error with p95 latency under 3ms.
On the AI side, I engineered RAG-grounded reply generation with Gemini and pgvector — multi-format knowledge-base ingestion, semantic retrieval with matched-chunk evidence, comment classification with confidence and sentiment, and post summarization — so responses stay on-brand and grounded in each customer’s own content.
I also designed the automation engine powering 1,000+ automations and 10,000+ concurrent conversations: branching multi-turn flows, follow/email gates, per-window cooldowns, and story-reply campaigns. To keep that configurable for non-technical users, I built visual flow editors and a Next.js dashboard for the DM tree, agent studio, and live analytics.
Beyond the core product, I implemented a Google Calendar booking engine that hands the AI opaque slot IDs (never raw times), creates idempotent bookings, and auto-DMs Meet links; shipped a full OAuth 2.1 authorization server (RFC 8414/9728, PKCE S256) and an MCP server so Claude/ChatGPT/Copilot can operate user accounts; and integrated Dodo subscriptions + AppSumo lifetime licensing through verified webhooks.
This role has given me end-to-end ownership across distributed systems, AI pipelines, and product-facing dashboards — a mix I’ve found genuinely energizing.