I'm Divyanshu Singh, a final-year BCA student specializing in Artificial Intelligence & Data Science.

My main interest lies in building systems that sit between data science and software engineering. I don't want my work to stop at a trained model inside a notebook — I enjoy taking an idea through the whole pipeline:

Problem
   ↓
Data
   ↓
Analysis
   ↓
Machine Learning
   ↓
API / Application
   ↓
Deployment
   ↓
Monitoring & Improvement

I'm especially interested in Machine Learning, MLOps, Cloud Computing, and AI FinOps. I'm curious about what happens after a model is trained: how it is served, monitored, evaluated, scaled, secured, and eventually optimized for operational and infrastructure cost.

A few things that describe how I work:

• 🧠 I like understanding why a system works, not just getting it to run.
• 🔧 I learn best by building complete projects and debugging real problems.
• 📊 I enjoy working with data, SQL, analytics, and machine learning.
• 🚀 I prefer practical projects with a clear problem and measurable purpose.
• ☁️ I'm expanding my understanding of cloud and production ML workflows.
• 💰 I'm exploring AI FinOps / AI cost optimization as a long-term specialization.
• 📚 I'm continuously strengthening my fundamentals while experimenting with newer AI technologies.

I can turn a random real-world problem into a data science project before I even realize I've started one. 😄

These are the projects that best represent how I like to learn and build.

ML Risk Auditor is an AI-powered dataset auditing platform that analyzes uploaded datasets, identifies common machine-learning data risks, and provides preprocessing recommendations.

A large amount of ML work happens before model training. Missing values, duplicates, constant columns, identifier-like features, and high-cardinality features can create problems long before an algorithm is selected.

This project explores how that repetitive dataset-auditing process can be automated.

• 📂 Multi-format dataset loading
• 📊 Dataset structure and statistical summary
• 🔍 Automated data-quality checks
• ⚠️ Risk/severity classification
• 💡 Preprocessing recommendations
• 🤖 AI-powered Data Science Copilot
• 🧩 Modular project architecture

Python Pandas Streamlit Groq API

→ View ML Risk Auditor

Instead of immediately replacing a production model with a new candidate, this project implements a Champion–Challenger workflow.

The existing model acts as the Champion, while a new candidate model acts as the Challenger. The Challenger receives a controlled portion of traffic and is evaluated before it becomes the production model.

• 🔀 90/10 Champion–Challenger routing
• 🚀 FastAPI model serving
• 🧠 Logistic Regression Champion
• 🌲 Random Forest Challenger
• 🗄️ MySQL prediction logging
• 📈 Accuracy, precision, recall & F1 evaluation
• ✅ Promotion eligibility checks
• 💾 Champion backup
• ↩️ Rollback support
• 🔄 Model reload capability

This project represents my interest in production ML, where model accuracy is only one part of the problem.

→ View Champion–Challenger ML API

SentinelAI is one of my projects where I explored the transition from an AI idea into an application with an API layer, interface, containerization, and cloud deployment.

• AI application development
• API architecture
• Streamlit UI
• Dockerized deployment
• Cloud deployment
• Working with external AI APIs

Python FastAPI Streamlit Docker AWS

→ View SentinelAI

LeadScout explores how public business/location information can be collected, structured, stored, and converted into useful lead intelligence.

The project combines API-based data collection with a database and interactive interface.

• 🌍 OpenStreetMap + Overpass API
• 📊 Business data processing
• 🗄️ MySQL persistence
• 🔎 Opportunity detection
• 🧭 Location-aware results
• 📱 Streamlit interface

Python OpenStreetMap Overpass API MySQL Pandas Streamlit

→ View LeadScout

DealTruth is a Chrome extension designed to answer a simple question:

Is the current price actually a good deal?

Instead of relying only on advertised discount percentages, the project records observed prices and uses historical statistics to produce an explainable deal assessment.

• 🔎 Product & price extraction
• 💾 Local price-history storage
• 📊 Historical price statistics
• 🎯 Deal Score
• 📈 Price-history visualization
• 🧠 Explainable results
• ⚠️ Discount-analysis heuristics

JavaScript Chrome Extensions HTML CSS Chrome Storage API SVG

→ View DealTruth

Alongside my larger projects, I also work on focused projects to strengthen specific skills.

I naturally gravitate toward projects that combine several layers rather than just one library:

                REAL-WORLD PROBLEM
                       │
                       ▼
                DATA COLLECTION
                       │
                       ▼
                DATA ANALYSIS
                       │
                       ▼
              MACHINE LEARNING
                       │
                       ▼
                  API / APP
                       │
                       ▼
                 DEPLOYMENT
                       │
                       ▼
              MONITORING / COST
                       │
                       ▼
                  IMPROVEMENT

That is the direction I'm intentionally moving toward: end-to-end systems rather than isolated notebooks.

I'm currently investing most of my learning time into the areas below:

Deepening fundamentals, model evaluation, feature engineering, preprocessing, and practical ML workflows.

Learning how models move from experimentation into reproducible, deployable, maintainable systems.

Building hands-on familiarity with cloud infrastructure and understanding where AI workloads actually run.

Exploring the economics behind AI infrastructure — inference cost, resource utilization, scaling, optimization, and cost-aware architecture.

Strengthening SQL, data pipelines, APIs, storage, and the foundations needed to support production ML systems.

My learning journey has included hands-on work and programs across:

• IBM SkillsBuild
• Data Science & Machine Learning
• Cloud technologies
• Databricks ecosystem
• Applied AI / Generative AI learning
• Python & SQL
• MLOps and deployment concepts

I care more about what I can build with what I learn than collecting certificates for their own sake.

I enjoy learning by experimenting.

Sometimes that means building an ML application. Sometimes it's understanding a database query, deploying an API, exploring a cloud service, or taking an ordinary idea and asking:

"Can data or AI make this better?"

That curiosity is a big part of how I approach technology.

I'm continuing to build toward a profile that combines:

Data Science + Machine Learning + MLOps + Cloud + AI Cost Optimization

My goal is to become comfortable not only with building models, but also with the engineering and operational decisions required to run AI systems in the real world.