afsar = {
"role": "AI/ML Engineer",
"mission": "Build practical intelligence for real-world decisions.",
"working_on": ["Agentic AI", "Multimodal AI", "Remote Sensing"],
"based_in": "Indore, India"
}I build end-to-end AI systems across fraud detection, healthcare forecasting, sustainability, and decision support - from data pipelines and model evaluation to APIs, interfaces, and deployment.
|
Agentic Remote-Sensing AI Analyst Exploring a natural-language interface for satellite-image analysis - combining remote sensing, vision-language models, and agentic model/tool selection.
|
AI-Powered Agriculture Platform Building a practical, farmer-oriented AI platform around applied machine learning and decision support.
|
|
AI-Powered Mule Account Detection A bank-investigation desktop application with risk dashboards, transaction analytics, network visualization, and explainable risk insights. LightGBM + Isolation Forest across a 26-feature behavioral pipeline.
🏁 SquidHack 2026 Finalist - Rank 7/75 |
Healthcare ML End-to-end ML workflow using the Wisconsin Breast Cancer dataset, with comparative model training and a Flask prediction application.
|
|
Processed 8,450+ records into a Streamlit dashboard with KPI cards, drill-down filters, and 30-day forecasting for social-welfare decision support.
|
AI-based composting solution. Technical lead for a four-person team that built ML-backed real-time composting recommendations.
🏅 Indian Patent Holder - GreenLoop AI Composting System |
|
Data → signal EDA, feature engineering, forecasting, and statistical analysis. |
Signal → model Classical ML, deep learning, anomaly detection, and explainability. |
Model → product FastAPI, Flask, dashboards, databases, and deployment. |
2026 Machine Learning Intern, Unified Mentor Pvt. Ltd.
└─ Built 30-day healthcare-capacity forecasting models and a Power BI monitoring dashboard.
2026 SquidHack Finalist
└─ Ranked 7th of 75 teams with MuleGuard in a 36-hour hackathon.
Indian Patent Holder
└─ GreenLoop AI Composting System.
Deep Learning ──► Computer Vision ──► Multimodal AI ──► Agentic Systems ──► Production AI
▲ │
└──────────────────── evaluation • deployment • reliability ──────────────────────┘
Currently deepening work in vision-language models, remote sensing, tool-using agents, evaluation, and production deployment practices.