Deep | Portfolio

QuantumLogic Labs — AI-Powered ATS Internship @ Mphasis

Full-stack AI Applicant Tracking System with dual-parser resume extraction, hybrid candidate scoring using weighted skill taxonomy, semantic similarity, impact analysis, and experience relevance.

Tech Stack: FastAPI, React, PostgreSQL, Google OAuth2, JWT, Docker

End-to-end hiring workflow: recruiter chatbot, JD optimizer, interview scheduling, analytics dashboard, and immutable audit logging.
Recruiter intelligence layer: Strong Hire / Hold / Reject decisions, skill-gap analysis, resume quality scoring, and candidate deduplication.

Drishti — AI Bilingual Assistant for Visually Impaired Capstone

Multimodal AI assistive system integrating Whisper, GPT-4o-mini, OpenCV, and real-time audio feedback to help visually impaired users interact through voice commands and scene understanding.

Tech Stack: Python, Whisper, GPT-4o-mini, OpenCV, Raspberry Pi

Low-cost prototype (~₹7K) with contextual memory and bilingual interaction — comparable to commercial smart-glasses priced ₹1.5L–3.5L.

NLP Text Summarizer with MLOps

End-to-end text summarization pipeline using the Pegasus Transformer with automated data ingestion, validation, model training, and experiment tracking via DVC and MLflow.

Tech Stack: Python, Pegasus Transformer, FastAPI, Docker, DVC, MLflow

FastAPI backend for real-time summarization inference and pipeline validation through APIs.

View on GitHub

E-commerce Sales Analytics Dashboard

Interactive Power BI dashboard analysing YTD sales, profit margins, category and regional performance from a 100K+ record dataset. Data modelling and transformation were implemented in MS SQL Server and Power Query to support performant visuals and dynamic KPIs.

Tech Stack: Power BI, MS SQL Server, DAX, Power Query, Data Modeling

Built dynamic KPIs, trend visuals (YoY), Top/Bottom product analysis and state-wise insights.
Integrated large datasets with ETL and optimized visuals for executive-level reporting.

View on GitHub

End-to-End Medical AI Chatbot

RAG-based medical assistant using The Gale Encyclopedia of Medicine as a domain-specific knowledge base. Full pipeline: text extraction, chunking, semantic embeddings, Pinecone vector store, and LangChain + LLM integration.

Tech Stack: Python, LangChain, Hugging Face, Pinecone, FLAN-T5, Flask

Flask web app for real-time context-driven health Q&A with grounded LLM responses.

View on GitHub

Smart Fitness Insights

AI-powered fitness dashboard built with R Shiny to monitor health metrics, analyze activity patterns, and deliver personalized workout and lifestyle recommendations.

Tech Stack: R, Shiny, ggplot2, Plotly, Random Forest, Isolation Forest

View on GitHub

Blood Management System

University-focused blood donor-recipient matching platform with secure role-based access, email-driven requests and donor management features.

Tech Stack: HTML5, CSS3, JavaScript, Node.js, Express.js, MongoDB, Nodemailer

View on GitHub

Sentiment Analysis

Speech sentiment analysis and visualization using NLP techniques and audio transcription pipelines.

Tech Stack: Python, NLTK (VADER), Google Speech Recognition, Matplotlib

View on GitHub

Talk Lens Ongoing

AI-powered smart glasses that read and translate text aloud based on hand gestures — integrating computer vision, OCR and low-power IoT for accessibility.

Tech Stack: Python, MediaPipe, PaddleOCR, OpenCV, ESP32, Arducam

View on GitHub