AI Agent Designer Projects in GurugramAI Agent Designer Projects in GurugramAI Agent Tutor Mobile App
An AI personal tutor mobile application designed for learners of all ages, from students to professionals. The app provides personalized guidance, explanations, skill development, and interactive learning across multiple subjects and topics.
Key Features:
> AI tutor conversations for learning any topic
> Personalized learning paths based on user goals and skill level
> Support for academics, professional skills, coding, languages, and general knowledge
> AI explanations, summaries, and concept breakdowns
> Practice exercises, quizzes, and knowledge improvement tools
> Voice and text-based learning interactions
> Learning history and progress tracking
> Adaptive AI responses based on user behavior
Key Contributions:
-- Developed a cross-platform mobile learning experience
-- Integrated AI/LLM technology for real-time tutoring interactions
-- Implemented secure authentication and data management
-- Optimized performance and user experience across devices
Outcome:
Created an AI learning companion that enables users of all ages to learn, improve skills, and access personalized education anytime through a mobile-first experience. AI Viral Story & Cinematic Shorts Factory
AI Viral Story & Cinematic Shorts Factory is a futuristic AI-powered storytelling workflow built entirely inside Melius.
The project explores how AI can transform a single emotional idea into a complete viral-ready cinematic production pipeline for platforms like YouTube Shorts, TikTok, Instagram Reels, and Pinterest Video Pins.
Using interconnected AI agents and advanced visual workflow nodes, the system handles:
Trend research and viral analysis
Emotional story generation
Cinematic scene breakdowns
AI image prompting
AI animation direction
Camera movement planning
Voiceover scripting
Music and sound design
Thumbnail optimization
Captions and hashtag generation
Engagement prediction
Multi-platform export workflows
For the core demonstration, I created a 6-scene emotional rescue story following an abandoned puppy in the rain. The workflow visually demonstrates how a raw emotional concept evolves into a fully cinematic short-form narrative through interconnected AI systems.
The project was designed to feel like a next-generation AI filmmaking operating system — combining storytelling psychology, prompt engineering, cinematography logic, emotional optimization, and creator workflow automation into one connected visual canvas.
My goal was to explore the future of AI-native filmmaking and demonstrate how creators can generate production-ready emotional content in minutes instead of weeks.
Process:
Researched emotional viral storytelling formats
Designed a multi-agent cinematic workflow architecture
Built interconnected AI production nodes inside Melius
Generated a full 6-scene cinematic narrative
Created image and animation prompt systems
Added voiceover, music, and sound design logic
Built thumbnail, caption, and engagement optimization systems
Produced a cinematic walkthrough showcasing the workflow
Feedback on Melius:
Using Melius felt like directing an AI-powered creative studio visually instead of switching between disconnected tools. The node-based workflow made it easier to structure cinematic storytelling pipelines, iterate on ideas, and connect production systems together in a much more intuitive way.
LinkedIn Post:
https://www.linkedin.com/posts/buildwithlalit_meliuschallenge-ugcPost-7462439292988968960-xnV2 What It Is
SevaFlow is a civic complaint management system that lets Indian citizens file government complaints through Telegram without downloading any app or creating an account. The complaint gets automatically understood, routed, and tracked using AI.
How It Works End to End
A citizen sends a plain text message to the Telegram bot describing their problem. That message gets sent to Google Gemini with a carefully designed prompt at temperature 0.1, meaning the AI outputs consistent, deterministic JSON every time. Gemini extracts the issue type, location, responsible department, priority level, and generates a summary, all returning a confidence score between 0 and 1.
The routing engine then takes over. It applies priority override rules first, so words like "fire" or "emergency" always trigger urgent regardless of what the AI said. It maps the AI suggestion to a configured department, assigns an SLA deadline based on department and priority, and stores everything in SQLite. The citizen immediately receives a Telegram confirmation with their reference ID like SF1234, department name, priority, and expected response time.
The Admin Side
Government officials log into a dashboard at the FastAPI server. They can filter and sort complaints, view the full status history of each one showing who changed what and when, update the status with notes like "team dispatched", and trigger a Telegram notification back to the citizen automatically.
What Makes It Technically Interesting
The AI pipeline has a two layer fallback. If Gemini fails, keyword matching kicks in to identify the department. If that also fails, it routes to General Services with medium priority and confidence marked as 0.0 so admins know it needs manual review. Nothing gets lost.
The department configuration is fully data driven. Adding a new government department requires zero code changes, just a new entry in config.py (http://config.py) with keywords, SLA hours, and contact email. The system picks it up on restart.
The database tracks two separate tables: complaints with all AI output stored alongside the raw text, and status history with a complete changelog including timestamps and the identity of who made each change.
Why It Won
Most hackathon civic tech projects build a web form. SevaFlow used Telegram as the interface because that is where citizens already are, made the AI classification reliable enough to actually route correctly, and built the full government side too, not just the submission side. End to end in one system, deployable on a single lightweight server. Mainframe is a conceptual AI studio — brand system + full multi-page website, built entirely in Stitch. Landing page with animated states, five inner pages (Labs, Studio, Services, Openings), and a complete DESIGN.md (http://DESIGN.md) covering color, type, and layout. Monochromatic, editorial, cinematic. The kind of system that usually takes weeks — done in one session.
Started with brand tokens, used them as the foundation for every page. Iterated hover states and animations without leaving the design context. The generated DESIGN.md (http://DESIGN.md) became the actual handoff doc.
Fastest I've gone from concept to a coherent multi-page system. Want: auto-propagating token updates across pages, finer motion controls, native multi-page navigation. Otherwise — best design-gen tool I've used since Figma got auto-layout. I built a professional, end-to-end AI Receptionist system designed to automate clinic appointment management. This isn't just a chatbot; it's an AI Agent that can reason, use tools, and manage a live database autonomously.
Key Contributions:
Agentic Reasoning: Integrated CrewAI with Llama 3.3 (Groq) to enable the agent to understand complex user intents (Booking vs. Cancellation) and relative time (e.g., "next Tuesday at 3pm").
Autonomous Tool Use: Developed custom Python tools that allow the agent to verify real-time availability in a SQLite database and execute atomic transactions without human intervention.
High-Performance Backend: Built a robust API using FastAPI to handle asynchronous requests between the AI agent and the database.
Premium Dashboard: Designed a modern, Glassmorphic UI using Tailwind CSS that provides a real-time sync of the clinic’s schedule.
The Result:
A seamless, hands-free system that reduces administrative overhead by 100%, allowing clinic staff to focus on patients while the AI handles the entire scheduling lifecycle.
Tech Stack:
Python, CrewAI, Groq API, FastAPI, SQLite, Tailwind CSS
