whoami:
name: "Kuunal Mistry"
role: "AI/ML Engineer in Training | Full Stack Developer"
base: "Mumbai, India"
currently: "B.Tech (AI & ML), Atlas SkillTech University — 2024 to 2028"
focus:
- Designing and shipping end-to-end ML systems, from EDA to deployment
- Building recommendation engines and predictive analytics pipelines
- Full stack product engineering across Python, Flask, and modern web stacks
- Translating research-grade ML concepts into usable, production-shaped apps
philosophy: >
I care about systems that hold up outside a notebook —
clean data pipelines, reproducible models, and interfaces
people can actually use.Engineering focus: I build applied AI/ML systems end-to-end — from data collection and exploratory analysis through model training to a deployed, user-facing product. My work spans classification and recommendation systems, with a growing emphasis on full stack delivery so the models I build are never stranded in a notebook.
Open To:
Languages
Frontend
Backend & Databases
AI / ML & Data Tooling
Cloud, DevOps & Tooling
| Domain | Proficiency | Details |
|---|---|---|
| Classification Models |  |
Random Forest, Logistic Regression — applied to real-world NASA solar flare data |
| Recommendation Systems |  |
Content-based & collaborative filtering, cosine similarity, cold-start handling |
| Exploratory Data Analysis |  |
Pandas-driven EDA, trend visualization, data cleaning & preprocessing |
| Model Deployment |  |
Flask-based serving of trained ML models to web interfaces |
| Data Visualization & BI |  |
Power BI dashboards, Pandas/Matplotlib-based analytics reporting |
| Deep Learning Foundations |  |
TensorFlow fundamentals, building toward neural architectures |
🛰️ Solar Flare Prediction Website — Predictive Analytics Platform
A machine learning system trained on NASA solar activity data to predict solar flare occurrences, paired with a web interface for visualizing predictions and insights.
| Attribute | Detail |
|---|---|
| Stack | Python, Pandas, Scikit-learn, Google Colab, HTML/CSS |
| Scale | Full NASA space-weather dataset, end-to-end EDA-to-deployment pipeline |
| Performance | ~80% classification accuracy (Random Forest / Logistic Regression) |
| Security | No sensitive data handling; static prediction interface |
| Impact | Demonstrates applied ML on scientific time-series data with a usable front end |
| Repository | github.com/KuunalMistry/solar-flare-prediction |
Built to move beyond a pure notebook exercise: raw NASA data was cleaned and explored with Pandas, trends were visualized to guide feature selection, and classification models were benchmarked before being wrapped in a lightweight web interface so predictions are actually consumable, not just printed to a console.
🧭 India Travel Recommendation System — Hybrid Recommender Engine
A hybrid recommendation engine for Indian travel destinations, combining content-based and collaborative filtering to generate personalized, cold-start-resilient suggestions.
| Attribute | Detail |
|---|---|
| Stack | Python, Pandas, NumPy, Scikit-learn, Flask, Cosine Similarity |
| Scale | Multi-feature destination dataset (category, season, ratings, location) |
| Performance | 0.72 Precision@5, 0.80 Hit Rate |
| Security | Stateless recommendation API, no user PII stored |
| Impact | Solves cold-start with a weighted hybrid model; deployed via Flask |
| Repository | github.com/KuunalMistry/india-travel-recommender |
The core engineering challenge was the cold-start problem — new users or destinations with sparse interaction history. A weighted hybrid model blending content-based similarity with collaborative signals was designed and tuned, then served through a Flask API for real-time personalized recommendations.
🎓 Atlas Connect — Student Event & Networking Platform
A full stack platform for campus event discovery, registration, and networking, built to streamline student engagement at Atlas SkillTech University.
| Attribute | Detail |
|---|---|
| Stack | Python, SQL, HTML, CSS, JavaScript |
| Scale | Multi-user platform with authentication, event registry, and notifications |
| Performance | Real-time event registration and notification handling |
| Security | Login/authentication system with SQL-backed user management |
| Impact | Centralized campus event discovery, reducing fragmented communication |
| Repository | github.com/KuunalMistry/atlas-connect |
Designed as a full end-to-end product: a Python/SQL backend handles authentication, event registration, and notifications, while a hand-built HTML/CSS/JavaScript frontend delivers the user-facing experience — an early full-stack proof point ahead of deeper ML specialization.
Hackathon Finalist · IES MCRC Hackathon 3.0
Analytics · Data Storytelling · Digital Innovation
Competed as a finalist in a multi-track hackathon spanning analytics, data storytelling, and digital innovation, designing and presenting a data-driven solution under time constraints.
- Built and deployed working ML systems under competition constraints using Python, Flask, and Power BI
- Delivered end-to-end data storytelling — from raw data to a presentable analytical narrative
- Collaborated cross-functionally to ship a working prototype within the hackathon timeline
Python Flask Power BI Data Storytelling Machine Learning
| Recognition | Details |
|---|---|
| 🏆 Finalist — IES MCRC Hackathon 3.0 | Recognized across Analytics, Data Storytelling & Digital Innovation tracks |
| 📊 Applied ML Performance | Achieved ~80% accuracy on NASA solar flare classification task |
| 🚀 Deployed ML Systems | Shipped ML-backed applications using Python, Flask, and Power BI |
Python
Business Intelligence
current_focus:
learning:
- Deep Learning fundamentals with TensorFlow
- Distributed systems design for scalable ML
- Advanced full stack architecture (React + Flask/Node)
building:
- Refining the India Travel Recommendation System toward production
- Expanding Solar Flare Prediction into a real-time dashboard
- New full stack AI-powered product concepts
exploring:
- MLOps and model deployment pipelines
- Cloud-native ML serving (GCP / AWS)
- Open source AI tooling
open_to:
- AI/ML Internships
- Research collaborations
- Hackathons and applied ML competitions
"Models that stay in a notebook don't ship — I build the ones that do."