● Developed a cross-platform ecosystem (mobile, desktop, and Chrome extension) enabling seamless file and tab sharing between devices under the same Wi-Fi network — without relying on third-party apps like WhatsApp or Drive.
● Implemented real-time communication using Redis Pub/Sub and WebSockets for instant tab synchronization and file transfer.
● Designed secure drag-and-drop file sharing between laptop and mobile, with automatic downloads and progress tracking.
● Built mobile and desktop apps in React Native with responsive UI and native API integration for local device access.
● Engineered backend APIs using FastAPI and PostgreSQL to manage user sessions, file metadata, and transfer logs.
● Focused on privacy-first architecture — ensuring no external logins or data storage beyond local and peer devices.

● Curated and unified a multi-source dataset (~391K images) from CIFAKE, CelebDF v2, and Deepfake datasets for AI-generated vs. real image classification.
● Performed in-depth dataset analysis including image size, aspect ratio, RGB distributions, and t-SNE embeddings to detect feature overlaps and domain shifts.
● Standardized preprocessing (224×224 resizing, normalization, augmentation) achieving 100% label consistency across datasets.
● Trained a ResNet18 baseline model, attaining 93.26% in-domain accuracy (F1: 0.93) .
● Extracted and analyzed misclassified samples to identify dataset biases and edge-case patterns.
● Developed a Streamlit demo app for real-time inference, enabling image uploads with confidence-based predictions (“AI” vs. “Real”).
● Delivered a comprehensive 4-page evaluation report summarizing dataset rationale, EDA visualizations, and model performance insights.

● Built a production-grade URL shortener with user authentication (JWT), click tracking, and an analytics dashboard.
● Designed scalable backend APIs in FastAPI with PostgreSQL + Redis for caching and performance optimization.
● Developed an interactive React frontend with real-time metrics, link management, and protected routes.
● Implemented full-stack features end-to-end, securely integrating REST APIs via Axios.

● Built an autonomous trading agent with LangGraph to manage multi-cycle workflows including data fetching, technical analysis, and trade execution.
● Implemented a manual RSI strategy using NumPy & Pandas on real-time stock data (yfinance) to trigger rule-based buy/sell/hold signals.
● Integrated Alpaca’s paper trading API for dynamic position sizing (10% equity), achieving 30% profit ($30K) in backtesting.
● Designed a modular AgentState with Pydantic and built a Streamlit dashboard to visualize trading decisions, price trends, and performance.

● Implemented a low-latency LRU cache integrated into a multithreaded web crawler for optimized performance.
● Used std::mutex for thread synchronization, effectively avoiding race conditions and improving throughput.
● Strengthened expertise in system-level programming, concurrency, and performance-critical C++ design.

● Led a team of 4 to develop an NLP-powered summarization tool combining extractive (TF-IDF, Word2Vec) and
abstractive (T5 Transformer) techniques.
● Built preprocessing pipelines for PDF parsing, tokenization, lemmatization, and regex-based cleaning.
● Fine-tuned the T5 model for abstractive summarization and evaluated performance using ROUGE metrics.
● Achieved an “Outstanding” grade for innovation, accuracy, and deployment readiness.