Scroll
// About
whoami
~/tushar/about.md
$ cat about.md
▎
// Toolkit
Skills & Stack
Languages
PythonC++CJavaScriptSQL
AI / ML
Machine LearningDeep LearningComputer VisionNLPGenerative AILLMsScikit-learnOpenCVYOLOv8TF-IDFEnsemble MethodsFeature Engineering
LLM Engineering
LLaMA 3Prompt EngineeringGroq APIConversational AILLM Integration
Web / APIs
React.jsFlaskFastAPITailwind CSSREST APIsFirebaseWeb Speech API
MLOps / Tools
GitGitHubVercelRenderDockerAWSPandasNumPyCI/CDGitHub Actions
Core CS
Data Structures & AlgorithmsOOPDBMSOperating SystemsComputer Networks
// Journey
Experience
Vision AI Intern
Jul 2026 – Present@ VaultStack AI
- ▸Developed and integrated AI-powered computer vision features using Python and OpenCV, enabling real-time image and video analysis for practical AI applications.
- ▸Worked on model experimentation and evaluation by testing different approaches, analyzing results, and improving model performance for production-oriented computer vision workflows.
- ▸Built and integrated backend APIs to connect AI models with application services, ensuring smooth communication between computer vision components and the overall platform.
- ▸Contributed to production AI systems by debugging issues, optimizing workflows, and collaborating on reliable deployment and integration of machine learning features.
PythonOpenCVComputer VisionDeep LearningBackend APIs
Full Stack Development Intern
Jan 2025 – Feb 2025@ Ideal Medical Solutions
- ▸Developed GST-compliant invoicing and billing workflows using React.js, TypeScript, Firebase, and Firestore to support reliable day-to-day business operations.
- ▸Implemented reporting and data-management modules that helped organize billing information and provide users with clearer access to business records.
- ▸Improved the frontend interface by building responsive components, refining existing workflows, and creating a more consistent and user-friendly experience.
- ▸Collaborated with the development team throughout implementation and testing, resolving issues and delivering reliable features aligned with the application's requirements.
React.jsTypeScriptFirebaseFirestoreTailwind CSS
// Selected Work
Projects
AutiScan — ASD Detection & AI Therapy Platform
- ●Built an end-to-end AI platform for behavioral analysis and ASD screening using YOLOv8, OpenCV, Flask, React.js, and LLaMA 3.
- ●Implemented real-time computer vision and facial landmark analysis from live webcam input for behavioral pattern detection.
- ●Integrated LLaMA 3 (7B) through Groq API with prompt engineering and multi-turn contextual AI conversations.
- ●Connected AI models, REST APIs, frontend workflows, and analytics into a complete full-stack platform.
PythonYOLOv8OpenCVLLaMA 3NLPFlaskReact.jsGroq API
SafeHer AI — AI-Powered Women Safety Platform
- ●Built a full-stack women safety platform with SOS alerts, live location tracking, guardian monitoring, evidence vault, fake calls, and AI assistance.
- ●Developed secure backend services using Express.js, MongoDB, Firebase Authentication, JWT, and role-based access control.
- ●Integrated Socket.IO, Leaflet, and Cloudinary for real-time communication, location tracking, maps, and secure media management.
- ●Implemented security middleware, rate limiting, and scalable backend architecture with responsive React dashboards.
React.jsTypeScriptExpress.jsMongoDBFirebaseSocket.IOLeafletJWT
IPO Sense AI — AI-Powered IPO Analytics Platform
- ●Built an AI-powered investment analytics platform for IPO tracking, portfolio management, stock insights, and market intelligence.
- ●Unified live market data, IPO information, financial news, and Groq-powered AI insights into a single analytics platform.
- ●Developed interactive dashboards for portfolio analytics, watchlists, stock analysis, and data visualization using Recharts.
- ●Built secure backend services with Express.js, Firebase Authentication, REST APIs, and automated data-fetching pipelines.
React.jsTypeScriptExpress.jsFirebaseGroq APIRechartsREST APIs
NavAssist — Smart Navigation Assistant for Visually Impaired Users
- ●Built an AI-assisted navigation platform with hands-free voice interaction using the Web Speech API and an ML-based intent classifier.
- ●Developed a speech-to-intent pipeline that converts natural voice commands into real-time navigation actions.
- ●Implemented Flask and Firebase for authentication, session management, and synchronized application data.
- ●Deployed the platform using Vercel and GitHub Actions with a responsive and accessibility-focused interface.
PythonFlaskFirebaseWeb Speech APIMachine LearningREST APIs
Mystery Maze — Fantasy Puzzle & Riddle Platform
- ●Built a fantasy-themed interactive riddle and puzzle platform for Digi Week at J.C. Bose University of Science & Technology, YMCA, with a focus on engaging multi-level gameplay.
- ●Implemented team registration, location-based challenges, multiple gameplay levels, and interactive riddles to create a structured and progressive game experience.
- ●Developed real-time leaderboard and score-tracking functionality with Firebase Authentication and Firestore-backed sessions for reliable player and team progress management.
- ●Added automatic progress recovery and responsive frontend interactions so participants could continue their gameplay while maintaining their scores and completed challenges.
HTML5CSS3JavaScriptFirebase AuthCloud FirestoreVercel
InvoiceGenie — GST Invoice & Billing Management System
- ●Built a cloud-based GST invoicing and billing platform supporting GST tax invoices, proforma invoices, automated invoice numbering, and complete invoice history management.
- ●Implemented customer and product management workflows along with monthly GST reports, company settings, and structured billing data persistence using Firebase Firestore.
- ●Developed an interactive dashboard with billing analytics and organized business information to help users manage invoices, customers, products, and financial records efficiently.
- ●Designed a polished responsive interface using React.js, TypeScript, Vite, Tailwind CSS, and shadcn/ui with reusable components and dark/light theme support.
React.jsTypeScriptViteTailwind CSSFirebase Firestoreshadcn/ui
Fake Job Posting Detection System
- ●Built an AI-powered fake job detection system using TF-IDF, Random Forest, and Gradient Boosting on the EMSCAD dataset containing more than 17,880 job postings.
- ●Developed an end-to-end NLP pipeline covering text preprocessing, feature engineering, TF-IDF vectorization, and preparation of job-posting data for machine learning models.
- ●Addressed class imbalance using SMOTE and combined ensemble learning techniques to improve the reliability of fake-job classification across different posting patterns.
- ●Achieved 90%+ F1-score and deployed the trained prediction workflow through Flask REST APIs, enabling real-time classification of job postings through application requests.
PythonScikit-learnNLPTF-IDFRandom ForestGradient BoostingPandas
// Beyond Code
Leadership
Jul 2025 – Present
Senior Coordinator — University Computer Centre & Digital Affairs
J.C. Bose University
- ●Contributed to digital infrastructure initiatives serving 5,000+ students.
- ●Lead a technical committee modernizing university digital systems.
Nov 2023 – Jan 2026
Member — Youth Red Cross
J.C. Bose University
- ●Coordinate campus recruitment drives, liaising with 20+ industry partners.
- ●Deliver technical workshops on DSA, ML fundamentals, and web development.
Aug 2025 – Present
Class Representative
J.C. Bose University
- ●Elected academic representative for a 40-student cohort.
// Academia
Education
2023 – 2027
B.Tech, Computer Engineering
J.C. Bose University of Science & Technology, YMCA — Faridabad
Relevant Coursework
Data Structures & AlgorithmsMachine LearningIntelligent SystemsComputer VisionData MiningComputer NetworksSoft Computing
8.80
/ 10.0 CGPA
// Say Hello
Get in Touch
I'm actively looking for Software Development, Full-Stack, Backend, and AI Engineering internships. Drop a message or reach out directly — I'll get back to you soon.