KUNDHAVE S
Backend & AI engineer. Builds production-grade systems that survive failure, scale under load, and ship at velocity.
From architecture decisions to deployment, observability, and failure recovery — full ownership, no hand-offs.
ACTIVE
OPERATIONS
Not showcases. Working systems engineered to production standards — with real failure handling, observability, and architectural rigour.
CUSTOS
Cloud-Native Pre-Trade Risk & Regulatory Intelligence Platform
Architected and deployed a four-module event-driven RegTech platform on Microsoft Azure, enabling real-time trade validation, LLM-based regulatory rule extraction from uploaded PDFs, immutable audit logging, and ML-driven anomaly monitoring. Designed to handle high-frequency simulated trade event streams with low-latency rule evaluation using Redis caching; implemented shadow-mode anomaly detection combining Isolation Forest and PyTorch Autoencoder with alert streaming via Service Bus. Delivered a full-stack cloud application with CI/CD on Azure Static Web Apps enabling live interaction with validation outputs, rule lifecycle management, and anomaly alert dashboards.
FULLY DEPLOYED ON AZURE: Cloud-native full-stack application with robust CI/CD pipelines.
RELAY
Fault-Tolerant Distributed Webhook Orchestrator
Fault-tolerant event delivery system for asynchronous notification of external services across payment lifecycle events. Decoupled architecture with durable event persistence and instrumented delivery pipeline.
At-least-once delivery semantics with idempotency keys — zero duplicate payment events.
ENTITLED
Secure Privileged Access Management System
Security-focused PAM backend implementing role-based access control for sensitive financial vault data. Zero-trust architecture with just-in-time access provisioning.
Automatic access revocation — privilege escalation with zero standing permissions.
TECHTRENDGPT
RAG-Based Technology Intelligence Chatbot
Retrieval-Augmented Generation chatbot for real-time technology discourse and news analysis. Vector similarity search over indexed tech articles with streaming AI responses.
Context-aware responses grounded in recent tech articles — not just training data.
SYSTEM
STACK
Organized by system layer — from language primitives to cloud infrastructure.
PARALLEL
PROCESSES
Concurrent services — leadership and volunteering running alongside core engineering systems.
Technical Team Member (AI/ML)
General Manager
Co-Head — PR & Sponsorship
State Publicity Head
Joint Secretary
SYSTEMS
THINKING
Engineering principles that drive architecture decisions — not rules to follow, but instincts built through iteration.
FAULT ISOLATION
Design for failure, not against it. Every service boundary is a blast radius limiter. Dead-letter queues, circuit breakers, idempotency keys — not afterthoughts.
OBSERVABLE BY DEFAULT
If it runs in production and isn't instrumented, it doesn't exist. Prometheus metrics, structured logs, distributed traces — wired in from day one.
EVENT-DRIVEN FIRST
Decouple services through durable event streams. Async processing, at-least-once semantics, and backpressure handling as architectural primitives.
OWNERSHIP MENTALITY
Architecture → code → deploy → monitor → fix. No partial ownership. Comfortable holding the pager and the pull request simultaneously.
OPEN
CHANNEL
Backend & systems engineering — architecture to production.