Work

Back to Home Archive & Technical Case Studies 6 Projects & Experience

Creating next level software & systems

  • Real-Time Financial Analytics & Market Prediction Platform
    Live App GitHub Repo
    Spring Boot (Java) FastAPI (Python) React 19 / Vite Apache Kafka Apache Airflow PostgreSQL (Star Schema) Redis Caching WebSockets Docker Compose

    Technical Highlights & Architecture:

    • Designed a microservices architecture with a Spring Boot (Java) ingestion service, a FastAPI (Python) analytics service, and a React 19/Vite dashboard, communicating via Apache Kafka event streaming.
    • Built an automated data pipeline ingesting live stock prices every 30 seconds from the Yahoo Finance API, with producer-consumer Kafka streaming, broker-failure fallback logic, and raw JSON persistence.
    • Developed an Extract-Validate-Clean-Load (ETL) pipeline orchestrated with Apache Airflow, feeding a dimensional star-schema PostgreSQL warehouse optimized for financial time-series queries.
    • Implemented multiple forecasting models (Linear Regression, Prophet, LSTM) and a risk analytics module (Beta, Sharpe Ratio, VaR) with a model registry tracking RMSE, MAPE, and R² for model comparison.
    • Built a news sentiment-analysis pipeline with confidence scoring, and an AI Copilot enabling natural-language queries over market data.
    • Added Redis caching for high-performance query optimization, JWT authentication, live WebSocket price streaming, and Prometheus/Grafana monitoring, containerized with Docker Compose.
  • AI-Powered Meeting Minutes & Transcription Tool
    GitHub Repo
    Flutter & Dart FastAPI (Python) ChromaDB Vector Store LangChain Faster-Whisper (Offline STT) Gemini API & Groq PostgreSQL Docker

    Technical Highlights & Architecture:

    • Built a RAG-based chatbot vector store in Python (rag.py) using ChromaDB and LangChain, with fast-fallback logic to clean and route data reliably and prevent request timeouts.
    • Implemented an offline, local Faster-Whisper fallback in Python (transcribe.py) for reliable speech-to-text transcription without dependency on internet connectivity.
    • Built FastAPI backend services in Python with psycopg2 for PostgreSQL integration, handling data storage, retrieval, and normalization for meeting records.
    • Integrated the Gemini API and Groq for AI-powered summarization and NLP-driven data extraction from unstructured meeting transcripts.
    • Optimized backend CPU/GPU resource usage by designing client disconnection monitors in main.py, improving pipeline efficiency.
    • Implemented JWT and bcrypt-based security layers, and containerized services with Docker and Docker Compose.
    • Developed the Flutter frontend with Provider for state management, integrating Firebase Auth and Google Sign-In.
    • Programmed native mono WAV audio chunk merging on the client side (meeting_provider.dart) to support speaker diarization.
  • Farmers' Platform — MERN Stack (Aug 2025 – Nov 2025)
    Live App GitHub Repo
    MongoDB Atlas Express.js React.js Node.js Translation API JWT Authentication

    Technical Highlights & Architecture:

    • Built a full MERN stack platform with product listings, crop guidance features, and secure JWT-based authentication.
    • Developed REST APIs using Node.js and Express, and integrated MongoDB Atlas with optimized Mongoose schemas for scalable data storage.
    • Implemented multilingual support using a translation API, enabling real-time UI translation into multiple regional Indian languages.
  • Social Habit Tracking Application — Flutter + Supabase
    GitHub Repo
    Flutter & Dart Supabase PostgreSQL (RLS) SQLite (Offline DB) Firebase Cloud Messaging

    Technical Highlights & Architecture:

    • Built a Flutter/Dart app with Supabase (PostgreSQL, Auth, Realtime, Storage) and an offline-first architecture using SQLite with automatic synchronization.
    • Applied Repository and Service layer patterns with Row Level Security (RLS) and conflict resolution for reliable offline/online data sync.
    • Implemented Firebase Cloud Messaging and Supabase Realtime subscriptions for live cross-device notifications.
  • Full-Stack Expense Tracker & CSV Ingestion Pipeline
    GitHub Repo
    Next.js Node.js MongoDB ETL Workflow AI Categorization

    Technical Highlights & Architecture:

    • Built a full-stack personal finance tracker using Next.js, Node.js, and MongoDB to manage expenses, budgets, and subscriptions.
    • Designed a CSV ingestion pipeline with auto-categorization of bank transactions — an ETL-style workflow for cleaning and structuring raw financial data.
    • Integrated AI-powered insights to automatically summarize monthly spending patterns and generate personalized financial suggestions.
  • Flutter & Backend Developer Intern • K.J. Somaiya College of Engineering
    Python & FastAPI ChromaDB & LangChain Faster-Whisper PostgreSQL & psycopg2 Flutter Provider Docker Compose Speaker Diarization

    Comprehensive Responsibilities & Outcomes:

    • Built a RAG-based chatbot vector store in Python (rag.py) using ChromaDB and LangChain, with fast-fallback logic to clean and route data reliably and prevent request timeouts.
    • Implemented an offline, local Faster-Whisper fallback in Python (transcribe.py) for reliable speech-to-text transcription without dependency on internet connectivity.
    • Built FastAPI backend services in Python with psycopg2 for PostgreSQL integration, handling data storage, retrieval, and normalization for meeting records.
    • Integrated the Gemini API and Groq for AI-powered summarization and NLP-driven data extraction from unstructured meeting transcripts.
    • Optimized backend CPU/GPU resource usage by designing client disconnection monitors in main.py, improving pipeline efficiency.
    • Implemented JWT and bcrypt-based security layers to protect sensitive data, and containerized services with Docker and Docker Compose.