Prasoon Kandel

Built my Space Shooter Game in Pygame over a week! 🚀
Started with just a ship on screen, then added smooth movement and shooting. Enemies came next, followed by reload pickups for extra ammo. Scoring made it competitive, and a game-over screen tied it all together. Wrapped it up with clean, object-oriented code for smooth gameplay. Now it’s fast, fun, and ready to play! 🎮

Built my Space Shooter Game in Pygame over a week! 🚀
Started with just a ship on screen, then added smooth movement and shooting. Enemies came next, followed by reload pickups for extra ammo. Scoring made it competitive, and a game-over screen tied it all together. Wrapped it up with clean, object-oriented code for smooth gameplay. Now it’s fast, fun, and ready to play! 🎮

Devlog – Boba Drops Website
I created a clean and visually appealing website for the YSWS program Boba Drops. The color palette was carefully chosen to be perfectly balanced and eye-catching, while the typography follows key design principles for clarity and elegance. The result is a modern, well-structured site that presents content beautifully and is easy to navigate.

Devlog: Enhancing My Voice Assistant with DistilGPT

What’s New?

I’ve integrated the lightweight DistilGPT model from Hugging Face into my voice assistant to make its natural language understanding faster and more efficient, without putting too much load on my laptop.

What I Did

Dataset Creation: I created a custom dataset with example commands, conversations, and tasks that my assistant should handle.

Model Integration: I plugged DistilGPT into my assistant’s backend so it can generate and understand text more naturally.

Fine-Tuning Plan: I’m planning to fine-tune DistilGPT on my custom dataset to make its responses sharper and more relevant.

Current Status

I’m still grinding through the fine-tuning plan, testing different ideas and tweaking the setup to get the best performance. I haven’t pushed the updated project to GitHub yet — I want to get the model tuned properly first.

What’s Next?

Finish fine-tuning the DistilGPT model.

Test how well the assistant handles real commands.

Clean up the code and add proper docs.

Push everything to GitHub once it’s ready.

Challenges

I need to balance the fine-tuning so it doesn’t overuse my laptop’s resources while still improving the assistant’s quality. Also, making the dataset broad enough to cover real-world use cases takes time and testing.

Desk Buddy Voice Assistant — Devlog
Initial Concept:
I wanted to build a lightweight, desktop voice assistant that could listen to my commands, respond naturally, and perform useful tasks like opening websites or answering questions — all without heavy dependencies or complicated setups.

Tech Stack Choices:

Used Python 3.13 for latest features and compatibility.

Avoided PyAudio due to installation issues, and chose sounddevice for audio recording.

Used speech_recognition library with Google’s speech API to convert voice to text.

Integrated Google’s Gemini AI via the google-generativeai Python package to generate smart responses.

Added pyttsx3 for offline text-to-speech, making responses audible.

Built a simple yet sleek GUI using tkinter for easy start/stop control.

Challenges Faced:

Handling audio input without PyAudio required switching to sounddevice and converting the audio format properly for recognition.

Managing API quota limits from Google Gemini forced me to add graceful error handling and fallback responses.

Ensuring the UI remained responsive while processing voice commands meant running recognition and AI calls in a separate thread.

What’s Next:

Adding more command shortcuts for common tasks.

Improving error handling and user feedback.

Possibly integrating real-time info (weather, news) with external APIs.

Optimizing the audio recording duration to improve responsiveness.