Systems Engineering & AI Portfolio
An overview of our custom tools, open-source projects, and platform deployments built for enterprise scale and performance.
SAAN AI Shield Enterprise AI Gateway
How bridging the gap between DevOps and strategic leadership led to a 44% reduction in AI Usage Cost
Web Application & Blog
Production-grade Next.js 14 application that demonstrates the stack we recommend to clients. It features a CMS-driven blog via Sanity v3 with portable text rendering and syntax-highlighted code blocks, real-time comment threads and post likes via Supabase Realtime and Row Level Security, one-click OAuth via Supabase Auth (Google + GitHub), transactional email notifications via Resend, and automated ISR revalidation so content updates go live within 60 seconds of publishing. The animated hero uses Framer Motion with scroll-triggered reveals. Deployed on Vercel with automatic preview URLs per Git branch — enabling safe visual experimentation without touching production. Three UI themes built in parallel feature branches (Synapse Dark, Viktor Oddy, Nexora Light) with a single PR-merge deployment process.
AI Share Market Analysis Engine
Built to demonstrate how LLMs can make financial data analysis accessible without requiring users to read charts. The platform ingests real-time tick data from NSE and BSE, computes RSI, MACD, Bollinger Bands, and volume-weighted average price indicators, then feeds the combined signal set to Claude API to generate a plain-English analysis with a confidence-weighted recommendation. Results are streamed to a Next.js dashboard with real-time chart updates via Supabase Realtime. Designed for Indian retail investors — output is in INR, company names match Indian market conventions, and the system understands F&O expiry cycles. Built on FastAPI + WebSockets for the data pipeline, PostgreSQL for historical data, Redis for tick caching.
SaaN SRE Agent — Autonomous Incident Response
SaaN SRE Agent is a production-grade autonomous incident response system designed to replace the 3 AM on-call escalation. It operates in 9 sequential layers: metric ingestion from Prometheus and New Relic, Z-score anomaly detection, log correlation via OpenSearch, causal graph construction, root cause analysis powered by Claude API, action plan generation, confidence gate (threshold starts at 95%, relaxes as audit trail proves reliability), controlled execution via Kubernetes API and Helm, and full post-incident reporting. Every action is logged with reasoning, confidence score, and outcome — giving SRE teams a complete audit trail while progressively trusting the agent with more autonomy. Published on Medium with over 2,000 reads.
EEIP — Enterprise Experience Intelligence Platform
EEIP transforms how large IT services organizations understand their own workforce. Instead of relying on outdated spreadsheets or expensive HR systems, EEIP continuously maps employee skills from delivery metadata — project histories, tech stacks, client domains — without ever touching personal or client-confidential data. The platform surfaces who knows what, who has delivered similar work before, and where capability gaps exist across the organization. Built on Next.js 14, FastAPI, PostgreSQL with pgvector for semantic skill similarity, Neo4j for relationship graphs, and Keycloak for enterprise auth with Active Directory integration.