Welcome to Folio, the portfolio of Syed Addan, a Machine Learning Engineer who loves blending creativity with cutting-edge technology. 🚀
Imagine a portfolio that talks. With Folio, you can ask the Voice AI anything about Syed’s work, skills, or projects—and it will answer truthfully, just like having a friendly conversation.
- A Conversational Portfolio: No need to read walls of text. Simply ask, and the Voice AI will do the rest!
- Smart Search: Whether you're curious about his resume or LinkedIn, Folio finds the right answers from the right place.
- Smooth & Interactive: Hover over buttons to preview resources or dive deeper by clicking. It’s simple, fun, and modern!
Because it proves Syed doesn't just understand technology—he makes it work in delightful, practical ways.
If you’re still scrolling, you’ve unlocked a little treasure: the "how" behind Folio. 🛠️
Folio is built to impress both technically and visually:
Built with Next.js, customized from LiveKit's WebRTC boilerplate. Features include:
- A clean, user-friendly interface.
- Smooth animations and hover previews for resources.
- Integration with the Voice AI for conversational interaction.
Folio is powered by two servers:
-
Application Server:
- Handles application logic and manages the context documents for answering user queries.
- Services include:
- LLM (Language Model): For generating conversational responses. Used: llama-3.1-8b-instant from groq.
- STT (Speech-to-Text): For processing user audio input. Used: nova-2-conversationalai from deepgram.
- TTS (Text-to-Speech): For generating voice responses. Used: eleven_turbo_v2_5 from elevenlabs.
- VAD (Voice Activity Detection): For detecting user speech. Used: silero-vad from silero.
- Turn Detection: To manage conversational flow. Used: turn_detector from livekit.
- RAG (Retrieval-Augmented Generation): For accurate, context-driven answers. Used: still in development.
- Web Searching: To fetch external information dynamically. Used: web search from exa.ai.
- Function Calls: For executing specific actions based on user queries. Used: inbuilt function calls from livekit.
- Hosting: Currently deployed on Microsoft Azure, but will migrate once free credits expire.
-
LiveKit Server:
- Leverages WebRTC for real-time audio and video communication.
- Hosted on LiveKit's cloud, ensuring low-latency communication.
- Integrated into the application using the LiveKit Python SDK.
- AI Pipeline: Speech-to-text, natural language processing, and text-to-speech combine for an engaging Voice AI experience.
- Search Intelligence: Contextual search uses LLMs to find relevant info from Syed's resume or LinkedIn.
- Animations: Dynamic previews and transitions keep the experience fresh and engaging.
If Folio inspired you to create something like it, here's a guide:
-
Set Up Servers:
- Use Python for your application server and deploy a LiveKit server for real-time audio/video.
-
Integrate AI Models:
- Choose models for STT, TTS, and LLM. Open-source models like Whisper, FastSpeech, and GPT-based ones are great starts.
-
Build the Client:
- Start with LiveKit's Next.js boilerplate, customize for smooth animations, previews, and navigation.
-
Deploy the Application:
- Containerize with Docker and host on a cloud provider (e.g., Azure, AWS, DigitalOcean).
-
Iterate and Innovate:
- Continuously improve your pipeline and client for a better user experience.
If you've scrolled this far, you're in the secret club. Remember, every line of code is a little step toward magic. 🚀
Folio is more than a portfolio—it’s Syed Addan's story, told in the most engaging way possible. Whether you’re here to hire or learn, welcome aboard! 🌟
If you’d like to connect: