Work That
Speaks Itself.

The client needed a high-volume core banking system capable of handling 50,000+ daily transactions with sub-10ms API latency, real-time ledger reconciliation, and full OWASP security compliance — all while integrating with their legacy SWIFT network.

We engineered a multi-tenant Next.js 15 frontend with a NestJS microservice gateway, PostgreSQL with row-level security, Redis session caching, and a WebSocket-based real-time transaction alert engine. The legacy SWIFT integration was wrapped in an async Celery queue to prevent blocking the main API thread.

The university had fragmented student information systems across 5 disconnected legacy databases. Faculty spent 12+ hours per week manually reconciling attendance, grades, and exam schedules across spreadsheets.

Built a unified SIS (Student Information System) with a Next.js frontend, NestJS REST API, and a unified PostgreSQL schema replacing all 5 legacy databases. Implemented automated grade computation, bulk attendance marking via QR scanner, and an exam conflict detection algorithm.

Managing client projects, invoices, meeting schedules, and developer task boards across 4 separate tools (Notion, Trello, Google Meet, Razorpay) was causing project delays, missed deadlines, and billing inconsistencies.

Designed and built the SkilloriaX unified platform combining a Client Dashboard, Developer Workspace, AI Requirement Analyzer, Razorpay billing pipeline, Jitsi video consultation rooms, and a real-time Socket.IO notification system — all in a single Next.js + NestJS monorepo.

AgriCorp operated 500+ agricultural delivery vehicles across 3 states with no real-time GPS visibility, leading to fuel theft, unauthorized route deviations, and a 23% on-time delivery failure rate.

Developed a Flutter cross-platform app with real-time Mapbox GPS tracking, automatic geofence alerts, driver attendance via biometric authentication, and fuel consumption analytics. The NestJS API integrates with an OBD2 vehicle telematics module for live engine diagnostics.

Elderly patients with chronic conditions were missing medication doses and experiencing health emergencies at home, with no real-time monitoring bridge between patients and their doctors between clinic visits.

Built a Flutter + BLE app that connects to Bluetooth health sensors (heart rate monitor, BP cuff, glucometer) and streams vitals in real-time to a doctor dashboard. An ML anomaly detection model triggers emergency alerts when vitals exceed threshold profiles.

Sales engineers spent 3–4 hours manually reading client project descriptions, estimating development costs, and writing technical proposals — a process prone to inconsistency and human error on 30% of projects.

Engineered a FastAPI + LangChain pipeline that parses natural language project requirements through a GPT-4 chain, extracts entities (features, complexity, integrations), maps them to our service catalog, generates itemized cost breakdowns, and produces a formatted PDF proposal — all in under 90 seconds.

LexAI's legal team manually searched through 50,000+ legal documents to answer client queries — a process taking 2–6 hours per query with inconsistent accuracy, causing expensive billing disputes.

Built a RAG (Retrieval Augmented Generation) pipeline using LangChain + Pinecone vector database. Legal documents were chunked, embedded with OpenAI embeddings, and indexed. A GPT-4 chain retrieves the top-5 semantically similar document chunks and synthesizes a cited answer with confidence scores.

HomeSmart needed a unified control platform for 50+ heterogeneous smart home devices (Philips Hue, Nest, custom ESP32 sensors) that currently required 7 separate apps — causing user frustration and preventing automation between device ecosystems.

Built an MQTT-based hub firmware on ESP32 that acts as a universal protocol bridge (Zigbee, Z-Wave, WiFi, BLE). The Node.js backend subscribes to all device topics and exposes a unified REST + WebSocket API consumed by a React dashboard and companion Flutter mobile app.

PharmaCold stores temperature-sensitive pharmaceutical products across a 200,000 sq.ft warehouse with 200 temperature zones. Manual spot-checks every 4 hours were missing critical temperature excursions, risking ₹2Cr+ in product losses per incident.

Deployed 200 ESP32 LoRa sensor nodes across all zones, each transmitting readings every 30 seconds to a central LoRa gateway. An InfluxDB time-series database stores all readings. A Grafana dashboard displays live heatmaps. Automated WhatsApp + email alerts trigger within 500ms of any zone exceeding threshold.