Welcome to my digital workspace!
I’m an AI Engineer passionate about building intelligent systems that empower real-world applications. From natural language agents to secure personal AI companions and full-stack automation, I build solutions that blend deep learning with clean user experience.
- 🎓 B.Tech in Artificial Intelligence & Data Science
- 🌱 Constantly exploring AI + App integration (Flutter + LangChain + Supabase)
- 🚀 Currently building personal assistant systems and private AI agents
- 💬 Ask me about LangChain, Flutter, AI chatbots, and FastAPI
- 🔭 I’m working on AI-enhanced apps that focus on privacy and personalization
- Python – AI, FastAPI, automation
- Dart / Flutter – Cross-platform mobile & desktop
- SQL – Supabase, PostgreSQL
- JavaScript Basics
- LLM Integration: LangChain, OpenAI, GroqAI, LLaMA 3
- Prompt Engineering, Fine-tuning, RAG pipelines
- Human-in-the-loop Agents
- Speech and Vision: Google Speech-to-Text, ML Kit (Face Detection)
- Object Tracking: YOLO
- NLP
- FastAPI, Flask
- Supabase (Auth, Storage, PostgreSQL)
- Google Cloud Run, Docker
- Zoho Creator (Low-code business apps)
- Zoho Creator: Automated business apps and workflows
- Zoho Books Integration: API-based sync between Creator and Books
- Deluge Scripting: Complex form automation, lookup filters, and UI logic
- Inventory & Billing Modules: Real-time stock sync, vendor validation, and dynamic GST calculations
- Git & GitHub
- Streamlit
- PowerBI & Tableau
- VS Code, Postman
A real-time vision system that sees and identifies objects in videos—like a search engine for the physical world.
- Analyze videos from local files, live RTSP streams, or YouTube links.
- Uses the YOLOv8 model for high-speed, accurate object detection and tracking.
- Users can visualize identified objects with bounding boxes and track them across frames in real-time.
- Ideal for automated surveillance, traffic analysis, and interactive video analytics.
- Built with Streamlit for an interactive web interface and OpenCV for video processing.
An advanced NLP tool that reads the emotional tone of any text—like an instant emotion detector for words.
- Paste text or documents into the app for immediate sentiment classification.
- Uses state-of-the-art Hugging Face Transformers to analyze sentiment in over 20 languages.
- Users receive instant feedback on whether the sentiment is positive, negative, or neutral.
- Perfect for social media monitoring, customer feedback analysis, and market research.
- Built with Flask to serve the model via a robust backend and a clean HTML/CSS frontend.
🔹 Baymax
A cross-platform personal companion that combines:
- Calendar + memory journaling
- Live user location & proximity
- GroqAI-powered assistant with task planning and personality
- Text vault for friends
Spiritual assistant with:
- Daily Quranic verse
- Salah tracker with analytics
- AI-based Islamic Q&A assistant (Noor)
- Natural language to SQL agent
- Supabase + LangChain + FastAPI
- Role-based access (admin vs employee)
- Realtime SQL generation and output
- Predicts CVD and Diabetes risk for Indian population
- Integrates LLama3 with custom prompts and mitigation plans
An intelligent system that allows users to search for specific words inside audio content—just like Ctrl+F for sound.
- Upload audio or provide links to long conversations, interviews, or lectures.
- Uses Google Speech-to-Text for transcription and Gentle Forced Aligner for precise word-level timestamps.
- Users can search for words, and the system jumps directly to that part in the audio.
- Ideal for journalists, researchers, and accessibility tools.
- Built using Docker, Streamlit, and pre-trained models for quick deployment.
- Email: [email protected]
- LinkedIn: Afthab Roshan
“Code like it matters. Build like someone will use it. Design like it tells a story.”