Hey everyone,
I just wanted to share a little about what I have been working on recently. It’s been a crazy ride with my internship and also building my own AI assisted voice chatbot. Honestly, when I started, I didn’t even imagine I would reach this stage.
So the chatbot idea came because I always liked the concept of talking to tech like it’s a friend. We already type so much, so why not just speak and let AI understand us? I worked on making a simple voice chatbot that listens to your voice, understands the command, and gives a proper reply. The main thing was making it feel natural, not robotic. It was tricky because voice input has errors, background noise, accents and all, but slowly I managed to handle those with some tweaks. I also added a bit of personality so it doesn’t sound boring.
At the same time, I was also doing my internship, which gave me a lot of exposure. I had tasks where I learned not only coding but also how to manage time, deadlines and sometimes stress. During the internship, I completed 5 important tasks (I still feel proud about that). Each one was like a small project of its own, and step by step I got more confident. It was not just technical, but also about discipline.
Combining both experiences, I realized one thing – working on projects teaches you more than just reading or watching tutorials. When you build something, even if it’s small, you face real problems and that’s where learning happens.
During my internship at Oasis Infobyte, I got the chance to work on 5 really interesting tasks. Each project was different, and every time I learned something new about data, coding, and problem solving.
Here’s a quick look at what I did:
Iris Flower Classification – my first step into machine learning, where I trained a model to classify flowers based on their features. It was a small project but gave me confidence to handle datasets and understand ML basics.
Unemployment Analysis – here I worked with real-world data to analyze unemployment trends. This task taught me how to clean data, visualize it, and get insights that actually make sense in real life.
Car Price Prediction – one of my favorites, where I built a model to predict car prices based on features like age, mileage, and brand. It made me realize how machine learning can be applied in business use cases.
Email Spam Detection – in this project, I created a classifier that can identify spam emails. It was exciting because it’s something we actually face in daily life, and I learned how natural language processing (NLP) works.
Sales Prediction – the final task, where I predicted future sales using data trends. This taught me how forecasting works and why companies depend so much on predictive models for decision-making.
Completing these 5 tasks gave me not only technical skills in Python, ML, and data handling, but also patience, confidence, and the ability to approach problems step by step.
Thanks for reading
Top comments (1)
Loved this write-up - you can feel the growth from juggling the internship and the voice chatbot. Making speech feel natural is way harder than it looks, especially with noise and accents.
A couple tips that helped me: try endpointing with both VAD + semantic cues (e.g., pause length + intent confidence) to reduce over-talk, and cache TTS snippets for common confirmations to hide latency. Also, do a pass with domain-specific lexicons to nudge ASR toward expected entities (names, brands, commands) - tiny dictionary boosts can clean up a lot of mishears.
I work on Fluents, where we build AI voice agents, and the “personality without sounding robotic” point really resonates. We’ve found a small set of conversational guardrails (ack + clarify + action) keeps flows natural while avoiding loops.
Curious: how did you evaluate “naturalness” during testing - subjective listening, user surveys, or any metrics like interruption rate or repair turns? Keep going, and congrats on knocking out those 5 projects.