Contract-driven, agentic system for deterministic CAD generation from natural language using headless CAD

BetWork is a two-stage project focused on designing a robust data pipeline and performing comprehensive data analytics for a betting platform. The project analyzes player behavior, betting patterns, and financial trends to extract valuable business insights, empowering decision-making and enhancing operational strategies.

Implementation of an ARM processor with hazard and forwarding units, along with SRAM and cache memory

Production-style NBA data pipeline that builds a partitioned BigQuery feature store from NBA Stats API endpoints for analytics and modeling workflows.

🚀 Convert natural language to real CAD artifacts with Text2CAD, an agent-driven system for deterministic and validated engineering workflows.

5 stages in-order pipeline architecture simulator

This project aims to create a pipeline-architecture for applying sentiment analysis to reviews on an amazon dataset.

CSE Bubble developed as semester project for Computer Organisation (CS220) course in 4th semester

Ingestion pipelines with artifact lineage, replayable stages, and append-only persistence.

Pipeline-driven job & internship alert system that aggregates, filters, and delivers relevant opportunities.

Real-time fraud detection pipeline (Airflow, SageMaker, PostgreSQL, Streamlit)

Enterprise-grade ML framework showcasing advanced Scikit-Learn implementations with production-ready pipelines, algorithm-optimized synthetic data generation, comprehensive evaluation suite with statistical testing, custom transformers, ensemble methods, and real-world industry applications across healthcare, finance, and manufacturing domains.

Structured data-cleaning pipeline built on the Olist e-commerce dataset, focusing on reproducible ingestion, schema profiling, and light standardization using Python and pandas. Emphasizes clear pipeline stages, auditability, and portfolio-ready documentation.

Semi-automated logo discovery pipeline built with Django and Python for extracting and ranking potential company logos from websites.

Content → Schema → Template → Format. Deterministic document transformation with clear separation between content, structure, and presentation.

A minimal Python starter for building repeatable data pipelines with optional AI/LLM processing. No frameworks — just clean, extensible structure.

A real-time Hindi → English voice translation pipeline with streaming speech recognition and offline ASR. Hindi-to-English speech translation system combining ASR (Kaldi/Vosk) with a transformer-based machine translation model, forming a complete speech → text → translation pipeline.

Deterministic first. LLM only where confidence is low. Built to reduce cost and latency, not maximize model usage.