Founder & Solo Developer
AI-focused Computer Science undergraduate and solo SaaS founder with production experience building real-time voice pipelines, full-stack web platforms, and native desktop applications. Seeking SWE and AI engineering internship roles.
GhostDesk
Feb 2026 – Present
Built a cross-platform Electron app for Windows and macOS with real-time AI coaching, voice transcription, OCR, and 14 interview modes. Engineered a dual-mode Deepgram STT pipeline with approximately 300ms latency and a hallucination filter that reduced transcription errors by approximately 90%. Deployed a serverless Vercel backend with 11 API endpoints, Google OAuth, Razorpay and DodoPay integration across four pricing tiers, automated device-locked entitlement management, multi-model AI streaming, native C++/Objective-C++ addons, AES-256-GCM storage encryption, and GitHub Actions CI/CD.
GeeksForGeeks Campus Body, Karnavati University
Aug 2025 – Jan 2026
Led technical initiatives and organized coding-focused events and workshops for more than 100 students. Mentored students in data structures, algorithms, and competitive programming strategies.
Developer Club, UIT Karnavati University
Dec 2024 – Jan 2026
Conducted structured data structures and competitive programming sessions for more than 30 juniors and peers. Developed and deployed the official Developer Club website and managed technical event logistics.
B.Tech, Computer Science Engineering (Artificial Intelligence)
Aug 2023 – Present
CGPA: 7.3 / 10
Offline real-time AI surveillance system built on consumer hardware with zero cloud dependencies. Implemented YOLOv8 object detection, persistent tracking, ONNX facial recognition, MediaPipe skeletal pose analysis, multi-threaded real-time inference, Ollama local LLM scene summarization, ChromaDB event memory and RAG chat, restricted zones, density heatmaps, and anomaly screenshots. Technologies: Python, YOLOv8, MediaPipe, OpenCV, ChromaDB, Ollama.
AI-powered competitive programming platform with personalized practice and performance tracking, growing to more than 400 active users. Built a multi-language online compiler for C++, Python, and Java, an adaptive Virtual Gym across 10+ difficulty tiers, an Llama-70B AI assistant, and a scalable Node.js/PostgreSQL backend. Technologies: Node.js, PostgreSQL, Express, React, Tailwind CSS, Docker, Redis.