Sanket Talaviya

AI-Powered Multi-Tenant Outreach Platform

AI Outreach Platform

Problem

Sales and outreach teams needed a scalable platform to manage multi-channel outreach campaigns with AI-generated messaging, but existing solutions lacked proper tenant isolation, conversation threading, and intelligent reply handling.

Role

Owned the full-stack architecture, designed the multi-tenant data model, built the AI integration pipeline, and implemented the real-time conversation system.

Architecture

Multi-tenant architecture with workspace-level database isolation, conversation-centric data model, thread-based messaging with AI reply parsing, and real-time engagement tracking via WebSockets.

Key Challenge

Building a true conversation-centric architecture that could handle complex threading across multiple channels while maintaining tenant isolation and supporting AI-generated content at scale.

Multi-Channel Product Information Management Platform

Categra

Problem

E-commerce businesses struggled with managing product data across multiple marketplaces. Catalog synchronization to Amazon took 12+ hours, product validation was error-prone, and managing parent-child product relationships was complex.

Role

Owned the catalog synchronization system, product validation engine, marketplace integration layer, and AI content generation features.

Architecture

Event-driven architecture with a rule-based validation engine supporting tens of thousands of business rules, marketplace-specific data transformers, and a product merge system with recovery and rollback capabilities.

Key Challenge

Reducing Amazon catalog sync time from 12+ hours to under 2 hours required a fundamental redesign of the sync pipeline, including batched processing, delta detection, and marketplace-optimized data transformation.

Content Intelligence & Recommendation Platform

Cachetag

Problem

Content teams needed a way to aggregate content from disparate sources (Dropbox, RSS feeds), organize it intelligently, and surface relevant content to users based on their behavior patterns.

Role

Architected the content ingestion pipeline, built the recommendation engine, implemented the Dropbox and RSS integration, and optimized SEO performance.

Architecture

Modular ingestion architecture with Dropbox synchronization and RSS feed parsing, a content recommendation engine based on user engagement signals, and an SEO-optimized content delivery layer.

Key Challenge

Building a recommendation engine that could process diverse content types and deliver personalized recommendations while maintaining high performance and SEO quality.

Industrial IoT Platform

IOTPOT

Problem

Industrial IoT deployments required a platform that could handle real-time telemetry from thousands of devices, maintain tenant-level data isolation, and provide live dashboards with RBAC.

Role

Designed the multi-tenant IoT architecture, implemented MQTT communication layer, built the Redis Pub/Sub and WebSocket real-time pipeline, and implemented RBAC.

Architecture

Tenant-isolated architecture with MQTT broker integration for device communication, Redis Pub/Sub for telemetry distribution, WebSocket streaming for real-time dashboards, and role-based access control for multi-user access.

Key Challenge

Ensuring reliable real-time telemetry processing across thousands of devices while maintaining strict tenant data isolation and low-latency dashboard updates.