Applied Info Solutions Pvt. Ltd. — Work Experience

Role: Machine Learning Intern Duration: 1 Month Employment Type: Internship Work Mode: In-Office

Overview

My time at Applied Info Solutions was brief — just one month — but it pulled me into a completely different world from mobile app development. Instead of building screens, navigation flows, and user interactions, I was working with models that had to make sense of raw, messy, real-world data: sensor coordinates, hand movements, and location signals — and turn that data into predictions that actually meant something to a real user.

It was a short internship, but a genuinely dense one. In the span of a month, I went from understanding how a machine learning model is trained, to actively contributing to a system that could recognize hand gestures, track a person's live location, and predict what they were likely to do next — all built around a single connected use case: an IoT-based geofencing system that could sense when a person moved outside a defined boundary, and alert them the moment it happened.

This document walks through that journey in the order I actually experienced it — starting from the fundamentals, moving into the specific techniques I was assigned, and ending with how everything came together into one working, predictive system.

Chapter 1: Understanding How a Machine Learning Model Actually Works

Before I could contribute to any real task, I had to understand what was actually happening under the hood when a model "learns" something — not as a buzzword, but as an actual process.

• I learned the end-to-end model training pipeline — how raw data gets prepared, cleaned, and fed into a model, how the model adjusts itself during training, and how its performance is then evaluated against new data
• I understood, at a practical level, how a model identifies patterns in data and uses those learned patterns to make predictions on inputs it has never seen before — the core idea that makes machine learning useful in the first place
• More importantly, I started to see ML not as an isolated script running in a notebook, but as one working component inside a larger system — something that had to take in real data, produce a real output, and hand that output off to the rest of the application to act on

This foundation mattered because everything that followed — clustering, KNN, gesture recognition, geofencing — only made sense once I understood this underlying loop: data in, pattern learned, prediction out.

Chapter 2: Learning Through Real Tasks — Clustering & K-Nearest Neighbours

With the fundamentals in place, I was assigned hands-on tasks built around two foundational machine learning techniques: Clustering and K-Nearest Neighbours (KNN). These weren't abstract exercises — they were the actual techniques used later in the project to make sense of location and movement data.

• Clustering: I worked with clustering algorithms to group similar data points together without any predefined labels — essentially letting the data reveal its own natural groupings. This gave me a real sense of how a model can find structure in data that, at first glance, looks completely unorganized.
• K-Nearest Neighbours (KNN): I implemented KNN to classify new, incoming data points based on how close they were to existing, already-known data points. It's a simple technique conceptually, but seeing it applied to real coordinate and movement data made it click in a way that reading about it never had — proximity-based reasoning turned out to be a surprisingly powerful way to predict behavior.

Together, these two techniques became the analytical backbone for the more applied, real-world tasks that came next.

Chapter 3: Teaching a System to Understand Human Movement — Hand Gesture Recognition

One of the more genuinely exciting parts of this internship was contributing to a hand gesture recognition model — a system designed to interpret physical human movement as meaningful, actionable input.

• I contributed to building and understanding a model capable of detecting and classifying different hand gestures, effectively teaching a system to "read" a human hand the way a UI reads a button tap
• This gave me direct, practical exposure to a fundamentally different kind of user interaction — one based on physical movement rather than a screen tap or click — and a real appreciation for how much harder it is to reliably interpret a gesture compared to a discrete, unambiguous button press

This piece of the project wasn't just a technical exercise — it was a glimpse into how ML-driven interfaces can let a system understand intent, not just input.

Chapter 4: Making Sense of Location — Geofencing & Coordinate Management

Alongside gesture recognition, a major part of the project involved working with geographic coordinate data to define and continuously monitor a virtual boundary — what's known as a geofence.

• I worked on geofencing coordinate management, handling the logic responsible for defining a specific geographic boundary and continuously comparing a person's live location against it
• I got direct exposure to how IoT connectivity was used to constantly stream live location data from a physical device into the system — data that then had to be processed, interpreted, and acted on in real time, not after the fact

This was the piece that made the project feel genuinely "real world" — it wasn't just data sitting in a file waiting to be analyzed; it was live, continuous, physical-world data that the system had to respond to as it arrived.

Chapter 5: Bringing It All Together — From Raw Data to a Real Alert

This is where everything I'd learned and built actually converged into a single, working, predictive system — and honestly, the most satisfying part of the internship.

• I contributed to integrating the trained model with the live IoT coordinate stream, so the system could take a person's current location and movement pattern and use it to predict their likely next move
• I helped build the logic responsible for detecting the exact moment a person crossed outside the defined geofence boundary, based purely on their real-time coordinate data
• Finally, I saw how the system would notify the user the instant they moved outside the permitted geographic area — closing the full loop: raw sensor data in, model prediction applied, and a real, meaningful, actionable alert delivered to the user

Seeing that full pipeline work — from a stream of raw coordinates all the way to a real notification appearing for a user — was the moment this internship stopped feeling like an academic exercise and started feeling like real engineering.

Skills Gained at Applied Info Solutions

Category Skills Machine Learning Fundamentals Model training pipeline, pattern-based prediction ML Algorithms Clustering, K-Nearest Neighbours (KNN) Applied ML Hand Gesture Recognition IoT & Location Intelligence Geofencing, Coordinate-based tracking, Real-time location monitoring System Integration Model + IoT data integration, real-time predictive alerts

Summary

Metric Detail Duration 1 Month Role Machine Learning Intern Work Mode In-Office Core Focus Clustering, KNN, Hand Gesture Recognition, Geofencing Project Type ML + IoT integrated predictive system

Even compressed into just one month, this internship gave me something valuable: a real, hands-on look at how machine learning moves beyond theory — from training a model, to feeding it live sensor data, to using its predictions to trigger genuine, real-world alerts for a person navigating physical space.