I build things that run on the web.

A CLI-first AI agent inspired by tools like ClawCode, built to handle tasks directly from the terminal while supporting different ways of working. Agent mode executes tasks, Plan mode structures the work first, and Ask mode keeps interactions focused on questions and explanations.

• Terminal-native AI agent interface
• Agent, Plan, and Ask interaction modes
• Task execution through the CLI
• Telegram interface for remote agent access
• Switch between interaction modes based on the task

A full-stack school management platform built for real-world deployment, connecting administrators, teachers, and students through role-based workflows — from school and class management to digital classrooms and teacher-created learning modules.

• Role-based admin, teacher, and student access
• School, class, section, subject, and classroom management
• Teacher-created modules and learning content
• Student access to classes and teacher content
• Administrative controls for managing the school

Generates a staged learning roadmap for a chosen career path, then produces the quizzes and counselling report for each stage. Built around a sequence rather than a chat box, so a learner always knows what comes next.

• Stage-wise roadmap generated per career path
• Quizzes and reports produced per completed stage
• Next.js frontend over a LangChain backend
• Progress carried across sessions

A RAG-based document Q&A system that lets users upload PDFs and ask questions grounded in their content, with relevant source passages returned alongside each answer.

• PDF upload and document chunking
• Embedding-based semantic retrieval
• Grounded answers using Groq-hosted LLMs
• Source passages shown with responses