I design and ship production-grade infrastructure — from real-time attendance engines with MQTT-driven RFID ingestion to multi-network Docker Compose deployments secured behind WireGuard tunnels. My flagship platform, SterlingONE, serves live at sterling.sabarikrishnan.me — a complete school/F&B management system built on Django 5, PostgreSQL 16, Redis 7, and EMQX 5.8, with 90+ API endpoints, Celery task pipelines, XGBoost predictive models, and Alpine.js real-time dashboards.
Beyond production systems, I architect neuro-symbolic AI assistants with hybrid RNN + LLM cores, RAG pipelines, and Model Context Protocol for persistent memory. Every system I build is latency-optimized, security-hardened, and designed to scale.
Deploying advanced logic into production environments.
Production · sterling.sabarikrishnan.me
A full-stack school & F&B management platform deployed on bare-metal infrastructure. Cortex is the AI subsystem powering predictive attendance analytics, anomaly detection, and intelligent automation across 70+ Edu API endpoints and 20+ F&B endpoints.
Architecture: Django 5 → PostgreSQL 16 → Redis 7 pub/sub → EMQX 5.8 (MQTT broker over WireGuard tunnel) → FastAPI ingestion → Celery task pipeline → Alpine.js real-time WebSocket dashboards. XGBoost models for risk-tier prediction with per-student what-if simulation.
Security: WireGuard-encrypted device mesh (10.0.0.0/24), Caddy TLS termination, EMQX bcrypt device auth, role-based unit scoping across 5 membership tiers.
ADVANCED NEURO-SYMBOLIC ASSISTANT
An extremely complex, latency-optimized intelligence system. Unlike standard assistants, Jarvis utilizes a hybrid neural architecture fusing Recurrent Neural Networks (RNNs) for sequential logic with Large Language Models (LLMs) for semantic understanding.
The infrastructure integrates Model Context Protocol (MCP) for persistent state management across sessions and utilizes RAG (Retrieval-Augmented Generation) to fetch real-time data, bypassing static knowledge cutoffs.
HEURISTIC SEARCH ALGORITHM VISUALIZER
A grid-based logic engine simulating high-stakes search protocols. Users deploy algorithmic strategies to locate targets within a hidden matrix, demonstrating A* pathfinding concepts and probability density.
Hover to focus on specific modules. Full production stack across backend, infrastructure, AI, and systems.