DEV Community: Raseena Anwar

Day 3: The Data Fuel – Structured vs. Unstructured, Labeled vs. Unlabeled

Raseena Anwar — Mon, 29 Jun 2026 13:48:49 +0000

Yesterday, we looked at how different algorithms learn. Today, we need to talk about the fuel that powers them: Data.

You’ve probably heard that "data is the new oil." But raw oil is useless until you know exactly what kind of engine you are pouring it into. In Machine Learning, the shape and state of your data determines your entire engineering roadmap.

Before writing a single line of model code, you have to look at your data through two critical lenses.

1️⃣ Structural Format: Structured vs. Unstructured Data

Before looking at what the data says, we have to look at how it is stored.
Structured Data (The Clean Spreadsheet): This is highly organized data that fits perfectly into traditional rows and columns (like a SQL database or an Excel sheet). Think of dates, phone numbers, transaction amounts, or inventory logs. Traditional ML models love this because it is incredibly easy to digest.

Unstructured Data (The Wild West): This is data that does not fit into a neat grid and it makes up about 80% of the world's enterprise data. Think of corporate emails, PDF legal contracts, customer service audio recordings, or video streams. Standard algorithms completely choke on this. To unlock it, you have to use Deep Learning architectures (like Neural Networks) that can read raw text tokens or images.

2️⃣ The Annotation State: Labeled vs. Unlabeled Data

Once you know the format, you have to check if it has an "answer key."

Labeled Data (The Ground Truth): This is data that has been paired with the exact target answer you want the model to learn. For example, a picture of a cargo truck tagged with the word "Truck", or a line of machinery metrics tagged "Failed". High-quality labeled data is gold, but it is incredibly expensive and time-consuming to create because it usually requires human effort to tag.

Unlabeled Data (The Raw Material): This is data in its natural, raw state no tags, no explanations, no answers. Think of a folder containing millions of customer comments or untagged receipts. It’s cheap and infinitely abundant, but the model has to figure out the patterns entirely on its own.

🎯 The Strategy Takeaway

When you sit down to solve a problem, look at the intersection of your data first:

Structured + Labeled: The gold standard for Supervised Learning. Perfect for predicting house prices or forecasting revenue.

Structured + Unlabeled: Direct path to Unsupervised Learning. Use this to find weird anomalies in server logs or discover natural customer segments.

Unstructured + Unlabeled: The foundational soil for Modern Generative AI.

This is how Large Language Models (LLMs) are builtby feeding them massive amounts of raw, untagged internet text so they can learn how language works on their own.

Before picking an algorithm, you have to audit your fuel.

AISeries #Day3 #DataEngineering #MachineLearning #DeepLearning #GenerativeAI #TechStrategy

Day 2: Supervised vs. Unsupervised vs. Reinforcement Learning

Raseena Anwar — Sat, 27 Jun 2026 09:34:46 +0000

When you pull back the hood of Machine Learning, you realize fast: algorithms don't all learn the same way. Just like humans, machines need different teaching methods depending on the problem at hand.

Strip away the math, and practical ML splits into three core paradigms:

Supervised, Unsupervised, and Reinforcement Learning.

Let's break them down with real-world scenarios instead of equations.

1️⃣ Supervised Learning - The Teacher-Led Classroom

Learning with a labeled dataset. You give the algorithm both the questions and the correct answers during training, so it learns the pattern linking them.

Use case: Predicting house prices.

Hand the model 10,000 past home sales square footage, bedrooms, location each paired with its final sale price (the "label"). It learns how much a bedroom or a zip code is worth. List a new house, and it estimates the market value based on what it learned.

2️⃣ Unsupervised Learning -The Independent Detective

Learning from unlabeled data. No answer key, no predefined categories. You hand over a pile of data and say: "Find the hidden structure for me."

Use case: Customer segmentation for e-commerce.

Feed the algorithm raw behavioral data - browsing time, purchase history, click patterns - across millions of shoppers. You never tell it what to look for. It might surface a cluster of "midnight impulse buyers" and another of "weekend discount researchers" - groups you never defined, discovered purely from behavior.

3️⃣ Reinforcement Learning -The Trial-and-Error Video Game

Learning through consequences. No historical dataset —an agent acts inside an environment and learns from rewards and penalties.

Use case: Training a robotic arm to navigate a warehouse.

You don't program exact joint physics. The robot tries to move. A clean step forward earns +1. A crash or dropped package costs -1. It fails constantly at first but after millions of simulated attempts, it converges on the movement sequence that works.

The Strategy Takeaway

Picking the wrong paradigm changes your entire engineering roadmap:

Forecasting an outcome from historical data (stock trends, spam detection) → Supervised

Exploring data for anomalies or natural groupings → Unsupervised

Building a system that makes sequential decisions in a dynamic environment (autonomous driving, game AI) → Reinforcement

Which paradigm is your current project living in?

And if you've ever had to pivot a project from one to another mid-stream I want to hear that story in the comments. 👇

Day 3 is next

AISeries #Day2 #MachineLearning #DataScience #SupervisedLearning #UnsupervisedLearning #ReinforcementLearning #TechStrategy

Day 1 AI Series: Demystifying the Spectrum - AI vs. ML vs. DL vs. GenAI vs. Agentic AI

Raseena Anwar — Fri, 26 Jun 2026 12:58:18 +0000

Kicking off a new series where I break down the AI landscape, one concept at a time. First up: the term everyone throws around but few define clearly.

If you're stepping into the tech world right now, the buzzwords can feel like alphabet soup.

"Isn't Generative AI just Machine Learning? Where does Agentic AI even fit?"

Let's cut through the marketing fluff with one mental model: nested concentric circles. Each layer is a specialized subset of the one surrounding it narrower in scope, but more powerful in what it can do
.
Here's the spectrum, from the outermost ring to the core.

🟢 1. Artificial Intelligence (AI)
The umbrella term for any technique that lets machines mimic human intelligence, logic, or behavior. This spans everything from 1990s rule-based chess engines to modern automation. If a system mimics a smart human decision, it's AI.

🔵 2. Machine Learning (ML)

A subset of AI where we stop hand-writing rules and let the system learn them from data instead. Feed an algorithm thousands of historical real-estate listings and sale prices, and it builds its own model to predict future prices.

🟣 3. Deep Learning (DL)
A specialized subset of ML modeled loosely on the human brain. Multi-layered neural networks automatically extract features from raw, messy data video, audio, images without an engineer manually labeling every feature by hand.

🟡 4. Generative AI (GenAI)
A specific capability within Deep Learning. Traditional ML/DL predicts or classifies ("Is this transaction fraudulent?"). GenAI creates something new. Powered by LLMs and foundation models, it takes a natural-language prompt and generates fresh text, code, images, or synthetic data.

🔥 5. Agentic AI - The Core
The current frontier. If GenAI is a brilliant conversationalist that responds when prompted, Agentic AI is an autonomous executor. These agents use an LLM as their reasoning engine, but pair it with tools, memory, and feedback loops. Give one a goal "Audit our cloud spend, flag the top three anomalies, and draft an optimization email to engineering" and it plans the steps, calls the right APIs, checks its own work, and runs the workflow end to end.

🎯 The Practical Takeaway
Matching the problem to the right layer saves time, budget, and compute:

Forecasting inventory or churn from spreadsheets? → Traditional ML

Summarizing a 50-page contract or drafting boilerplate code? → Generative AI

Running a multi-step workflow that pulls from databases and makes operational calls on its own? → Agentic AI

Where are you spending most of your time on this spectrum right now - ML, GenAI, or Agentic AI? Drop it in the comments. 👇

Follow along for Day 2, where we'll go one layer deeper.

AISeries #Day1 #ArtificialIntelligence #MachineLearning #DeepLearning #GenerativeAI #AgenticAI #TechStrategy

Journey of From Apron to Algorithms

Raseena Anwar — Thu, 02 Apr 2026 14:12:11 +0000

WeCoded 2026: Echoes of Experience 💜

Raseena Anwar

Mar 7

From Apron to Algorithm: Rebuilding My Tech Career After a Long Pause

#devchallenge #wecoded #dei #career

3 min read

From Apron to Algorithm: Rebuilding My Tech Career After a Long Pause

Raseena Anwar — Sat, 07 Mar 2026 17:13:01 +0000

This is a submission for the 2026 WeCoded Challenge: Echoes of Experience

There was a time I stopped calling myself an engineer.

Not because I forgot what I learned.
But because life slowly took me in another direction.

I graduated in Computer Science Engineering with dreams of building a career in tech. Like many others, I imagined myself growing in that world.

But my journey began somewhere else as a lecturer in India.

I loved teaching. Explaining concepts, helping students understand, watching them grow… it gave me purpose.

Then life changed.

After marriage, I moved to the UAE. Responsibilities grew. Family came first. I continued teaching for some time, but eventually, my career paused.

And when a career pauses, something else quietly begins doubt.

Technology doesn’t wait.

Every day there was something new. New tools, new frameworks, new ideas. And slowly, a thought started growing inside me:

“Maybe I’m too late.”

Years passed like that.

I didn’t say it out loud, but I felt it.

Like I had missed my moment.

One day, I opened my laptop again.

Not with confidence.
Not with a plan.

Just… with a small thought: let me try.

That was the hardest step.

Because starting again after a long gap doesn’t feel like continuing it feels like beginning from zero.

I had to relearn things I once studied.
I got stuck on simple errors.
I searched for answers that used to come easily.

Some days I felt proud.

Some days I questioned everything.

But slowly, something started changing.

Confusion became understanding.
Fear became curiosity.

I started building again.

Small projects first. Then bigger ones. I explored full stack development, tried new technologies, and slowly stepped into areas I once thought were out of reach.

But the most important part of this journey wasn’t the code.

It was the realization that I was not alone.

I started meeting women who had stories like mine careers paused for family, dreams quietly pushed aside, confidence slowly fading.

And I kept thinking…

Why do so many of us feel like we cannot come back?

That’s how Apron to Algorithm started.

Not as a big idea.

Just as a simple belief:

A woman who once paused her career can start again.

The apron and the algorithm are not opposites.

They are just different parts of the same story.

Today, I am still learning.

I build projects.
I share my journey.
I guide beginners when I can.
I try to create a space where others feel they can start again too.

I’m not chasing perfection anymore.

I’m just not stopping.

When I look back, my career doesn’t look like a straight line.

It moved through teaching, family, long pauses, self-doubt… and then a quiet restart.

But every phase gave me something.

Teaching gave me patience.
Motherhood gave me strength.
The gap gave me resilience.

And coming back gave me confidence I never had before.

In tech, we often celebrate fast success.

But not all journeys are fast.

Some happen late at night.
Some happen between responsibilities.
Some happen silently, without anyone noticing.

Until one day… they don’t.

If you feel like you are too late, or your career paused too long…

You are not behind.

You are just starting from a different place.

Some careers grow in a straight line.

Mine didn’t.

Mine paused, changed, struggled… and started again.

And sometimes, that restart becomes the strongest part of the story.

Sometimes the most powerful comeback doesn’t begin in an office.

It begins quietly… at a table, with a laptop open, and the courage to try again.

SafeSignal UAE

Raseena Anwar — Sun, 01 Mar 2026 10:35:23 +0000

This is a submission for the DEV Weekend Challenge: Community

The Community

For the last two days, we’ve all felt it.

The tension.
The uncertainty.
The constant refresh of news feeds about escalating war issues with Iran.

Even when you’re not directly involved, a regional conflict changes the air. It creates fear, confusion, and one dangerous thing — rumors.

That feeling inspired this project.

I built this for the 13 million residents and expats of the United Arab Emirates during a fictional March 2026 crisis scenario.

The UAE is one of the most diverse countries on Earth, home to over 200 nationalities. In moments of uncertainty, communities often retreat into language-specific WhatsApp groups — Arabic groups, Indian groups, Pakistani groups, Filipino groups, Western expat groups. And inside those spaces, information moves fast… but verification does not.

Rumors spread faster than facts.

And when panic spreads, people feel alone — even in a country of millions.

This community includes Indian and Pakistani expats, Filipino workers, Arab nationals, Western tourists — all living side by side, but often disconnected in crisis moments.

What they need isn’t more noise.

They need one calm, trusted voice.
One place that speaks their language.
One system that understands their exact neighborhood.
One signal they can rely on.

That’s why I built SafeSignal UAE.

What I Built

SafeSignal UAE is a community survival network that turns individual panic into collective power. It introduces NEIGHBOUR, a multilingual AI persona that acts as the "wisest person in the building."

Key Features:

Rumor Killer: Instantly fact-checks viral claims against official UAE sources (NCEMA, Dubai Police) to stop panic before it starts.
Community Board: A real-time marketplace for human kindness. It automatically extracts Offers (e.g., "extra water in Marina") and Needs (e.g., "insulin in Khalifa City") from chat and pins them by neighborhood.
Skill Finder: Quietly connects doctors, nurses, and first-aiders to those in their immediate vicinity.
Area Pulse: A ground-truth tracker that shows real-time safety reports from people standing in those exact locations.
Multilingual Core: Native support for Arabic, Hindi, Urdu, Tagalog, Malayalam, and English.

Demo

🔴 Live AI Studio App

https://ai.studio/apps/9f98c6bb-2cda-4bf7-9fef-f51585af843d?fullscreenApplet=true

💻 GitHub Repository

https://github.com/raseenaanwar/SAFESIGNAL-UAE

Code

The project is built with a modern, high-performance stack:
Frontend: React 18 + Vite
Styling: Tailwind CSS 4 (Custom "Military Ops" Design System)
AI Engine: Gemini 3.1 Pro (via @google/genai)
Real-time Intel: Google Search Grounding for live verification

How I Built It

The design philosophy was "Military Operations Center meets Human Safety." I wanted the UI to feel authoritative and high-trust, like a Bloomberg terminal, but with the warmth of a neighborly conversation.

The AI Brain: I used Gemini's advanced reasoning to create the Neighbour persona. It doesn't just provide data; it acknowledges human fear first. I implemented custom JSON extraction to turn raw chat text into structured "Community Posts" (Offers/Needs).

Real-time Verification: To solve the "hallucination" problem in a crisis, I integrated Google Search Grounding. Every safety tip or rumor check is verified against live news and official government feeds.

UI/UX: Built a 3-column dashboard featuring a live ticker for urgent updates, a "Live Pulse" sidebar for neighborhood status, and a "Quick Dial" card for emergency services tailored to the UAE's expat demographics.

Animations: Used CSS keyframes and motion to create pulsing status dots and sliding panels that communicate a "live" system state.

SafeSignal UAE proves that in a crisis, our greatest resource isn't just information—it's each other.

Career Comeback Coach: Turning the "Gap" into a Launchpad

Raseena Anwar — Sun, 01 Mar 2026 10:06:04 +0000

This is a submission for the Built with Google Gemini: Writing Challenge

opening a code editor after three years away felt like trying to speak a language I’d forgotten.

Whether it’s maternity leave, recovering from a health crisis, or taking time to care for a loved one, the "career gap" is often treated like a black hole on a resume. You look at a job description for a "Junior" role and realize the tech stack didn't even exist when you last worked. The industry moves like a high-speed train, and when you’ve been standing on the platform for years, trying to jump back on feels less like a career move and more like a leap of faith.

I built Career Comeback Coach because "just Google it" is terrible advice for someone dealing with the "imposter syndrome" that comes with a gap. You don't just need links; you need a bridge. You need someone—or something—to tell you that your time away wasn't "empty," and that your old skills still matter..

What I Built with Google Gemini

The app is a one-stop-shop for career returners. Instead of just a generic chatbot, I used Gemini 3 Pro to create a specific "Coach" workflow:

The Resume Reality Check (Vision): You upload a PDF or even a photo of your old resume. Gemini uses its vision capabilities to parse the layout and understand your history without me having to write complex OCR code.

Deep Reasoning (Thinking Mode): This is the "secret sauce." I used thinkingConfig to let Gemini "think" before it speaks. It looks at your old experience and finds hidden transferable skills—like how managing a household or a medical recovery actually translates to project management or resilience.

Real-Time Roadmaps (Grounding): AI loves to make up fake links. To fix this, I used Google Search Grounding. When the app suggests you learn "React 19" or "Next.js," it checks the web to give you actual, working documentation links from 2025.

The Voice Interview: Using the Web Speech API and Gemini, you can actually talk to the app. It acts like a supportive but firm hiring manager, giving you a safe space to practice the "So, tell me about this gap..." question.

Demo

Youtube Demo link

AI Studio Link :

accounts.google.com

What I Learned

Latency is the enemy of conversation.
When using "Thinking Mode," the model takes a few extra seconds to reason. I learned that you can't just leave a user staring at a blank screen. I had to build a "thought-trace" UI so the user can see what the AI is thinking about while they wait.
Designing for Empathy.
Technically, I learned a ton about React 19 and the Google GenAI SDK. But the real lesson was in the "soft" side. I had to prompt Gemini to be encouraging but not "fake." Finding that balance in the system instructions was harder than the actual coding.
Graceful Failures.
Sometimes APIs hit rate limits. I built a fallback system where if the "Deep Reasoning" model is busy, the app automatically switches to the standard Gemini 3 model so the user never sees a 500 error.

Google Gemini Feedback

The Good:
The Multimodal support is incredible. I didn't have to worry about file formats; I just sent the image/PDF and it "got it." Also, Search Grounding solved my biggest fear: giving a struggling student a 404 link.

The Bad:
The "Thinking Mode" is a bit of a double-edged sword. The output is 10x better, but the wait time can be tough for a "live" chat feel. I’d love to see a streaming version of the "thought process" so I can show it to the user in real-time.

The Alchemist’s Journal: An AI-Powered Portfolio Sanctuary

Raseena Anwar — Thu, 29 Jan 2026 14:00:33 +0000

This is a submission for the New Year, New You Portfolio Challenge Presented by Google AI

🌿 About Me
I am Raseena Anwar—a Software Engineer, AI Builder, and former Educator. After 7 years of teaching Computer Science, I transitioned into AI engineering with a mission to build systems that build people. My portfolio, "The Alchemist's Journal," is designed as an Atmospheric Sanctuary. In a world of high-speed scrolling, I wanted to create a space that feels like an editorial journal—calm, intentional, and deeply human.

🚀 Live Demo

🛠 The Architect's Journey: Building with Antigravity
This project was developed using Google Antigravity, the agent-first IDE. Moving beyond traditional coding, I acted as the Architect, orchestrating autonomous agents to handle the "gravity" of boilerplate so I could focus on the "lift" of creativity.

🌌 Agentic Workflow
Using Antigravity’s Manager View, I dispatched specialized agents to handle distinct parts of the sanctuary:

The Layout Agent: Managed the complex CSS Grid and Flexbox logic required for the "Digital Book" metaphor, ensuring the central spine remained responsive across all devices.

The Interaction Agent: Handled the Framer Motion orchestration for the "ink-line" micro-interactions. By using Antigravity's Implementation Plans, I could review and tweak the physics of the animations before the agent ever touched the main branch.

The Deployment Agent: Streamlined the containerization process for Google Cloud Run, ensuring that the artifact registry and service account permissions were perfectly configured for a seamless launch.

🧠 The Gemini Integration: Empathy-Driven AI
The heart of this portfolio is the AI Mentorship Playground, powered by Gemini 1.5 Flash via the Google Generative AI SDK.

I didn't want a "chatbot." I wanted a digital reflection of my teaching philosophy. Using a custom system instruction, I tuned the model to act as a mentor. Visitors can converse with this "AI Twin" to:

Contextualize my journey: Ask how 7 years of teaching informs my approach to React 19 or Next.js 15.

Interactive Bio: Instead of reading a wall of text, users can ask, "How does Raseena approach community building through code?"

Low Latency Experience: I chose Gemini 1.5 Flash for its near-instant response times, ensuring the dialogue feels as natural as a face-to-face conversation in a classroom.

✨ What I'm Most Proud Of
The "Digital Book" Metaphor: I’ve reimagined the portfolio as a journal. Every section is a "page," creating a structured rhythm that respects the reader's pace.

Micro-interactions: "Lines Becoming Meaning": When you hover over my name, an organic ink line draws across the screen, revealing core values: Resilience, Empathy, and Architect.

Elevated Through AI: By leveraging Antigravity and Gemini, I proved that AI is the ultimate collaborator. It didn't just help me code faster; it allowed me to build a more sophisticated, high-fidelity experience that truly represents my dual identity as a teacher and an engineer.

📂 Source Code
The complete source, including the Antigravity implementation plans and Gemini prompt engineering, is available here: GitHub:https://github.com/raseenaanwar/raseena-anwar-portfolio