┌─────────────────────────────────────────────────────────────────────┐
│                                                                     │
│   AI-Powered Platforms   →   PropMind, AI Workflow Automation       │
│   Distributed Systems    →   Kafka, BullMQ, Redis, Event-Driven     │
│   Production SaaS        →   LeaseHub, Multi-Tenant, Stripe         │
│   Cloud Security         →   AWS IAM Drift Detection, Lambda        │
│   ML Engineering         →   SHAP, Attrition Prediction, Explainable│
│                                                                     │
└─────────────────────────────────────────────────────────────────────┘

Business logic belongs in code — not spreadsheets, not reminders, not trust.
I design backend-first systems where payments are verifiable, workflows are automated, and failures are traceable.

Node.js Express MongoDB OpenAI API Stripe BullMQ Redis Socket.io

The problem: Property management is full of manual steps — approving applications, chasing rent, triaging maintenance. PropMind replaces every manual step with code.

sequenceDiagram
    participant T as Tenant
    participant S as System
    participant AI as AI Layer
    participant ST as Stripe
    participant A as Admin

    T->>S: Submit application
    S->>AI: Score tenant risk
    AI-->>S: Risk score + confidence + red flags
    S->>A: Notify with AI analysis
    A->>S: Approve application
    S->>T: Deposit required (timer starts)
    T->>ST: Pay deposit
    ST->>S: Webhook (signature verified)
    S->>S: Property → RESERVED, Ledger entry created
    Note over S: 1st of month — cron fires
    S->>S: Generate invoice for all ACTIVE leases
    S->>T: Rent reminder (3 days before due)
    T->>ST: Pay rent
    ST->>S: Webhook verified
    S->>S: Invoice → PAID, LedgerEntry, Notification
    A->>S: View AI financial summary
    S->>AI: Summarize revenue, dues, overdue
    AI-->>A: Natural language financial narrative

What's under the hood:

FastAPI PostgreSQL React LLM APIs Prompt Engineering

An orchestration engine that lets you define, schedule, and monitor AI-powered workflows without writing a new pipeline each time.

User defines workflow
        │
        ▼
┌──────────────────────┐
│  Workflow Scheduler  │  ← PostgreSQL-backed task queue
└──────────┬───────────┘
           │
    ┌──────▼──────┐
    │  Execution  │  ← Retry handling, observability
    │   Engine    │
    └──────┬──────┘
           │
    ┌──────▼──────────────┐
    │   LLM Integration   │  ← Prompt-driven decision nodes
    └──────┬──────────────┘
           │
    ┌──────▼──────┐
    │  Analytics  │  ← Pipeline metrics, execution history
    └─────────────┘

Node.js Express.js React MongoDB Docker Stripe

Production-grade SaaS. Multi-tenant isolation, Stripe payments, event-driven payment webhooks, full CI/CD. Zero manual invoicing.

Spring Boot Apache Kafka Redis PostgreSQL

Event-driven messaging pipeline with producer/consumer services, dead-letter queues, retry mechanisms, and fault-tolerant asynchronous processing at scale.

Producer Service  →  Kafka Topic  →  Consumer Service
                          │
                    Dead Letter Queue (on failure)
                          │
                    Retry Worker  →  Resolved / Alerting

AWS Lambda EventBridge CloudTrail DynamoDB

Real-time cloud security monitor. Detects privilege escalation and unexpected IAM changes across an AWS environment the moment they happen — not the next morning.

Python Scikit-learn SHAP Streamlit

End-to-end ML pipeline: preprocessing → feature engineering → model training → SHAP-based explainability → interactive Streamlit dashboard. Not just predictions — transparent reasoning.

const nandu = {
  focus:      ["backend-first", "systems thinking", "production-grade"],
  avoids:     ["CRUD apps with no real logic", "trust over automation"],
  believes:   "business rules should be enforced by code, not people",
  currentlyImproving: ["DSA patterns", "system design at scale", "distributed tracing"],
  openTo:     "SWE roles — backend, AI systems, platform engineering"
};