From Raw Data to
Deployed Intelligence
Machine learning engineer focused on reproducible pipelines, computer vision, NLP systems, and deployable APIs. Final year B.Tech AI & Data Science student building projects with the habits of a production ML team.
Production ML Work
End-to-end AI projects covering pipelines, model integration, deployment surfaces, and technical leadership.
Network Security Threat Detection Platform
PROBLEM: Network threat detection requires continuous, reliable classification of high-dimensional traffic data. Without validation and reproducible training, model quality degrades silently. SOLUTION: Engineered a modular ML pipeline covering MongoDB ingestion, schema validation, KNN-based imputation, drift checks, experiment tracking, model versioning, and FastAPI inference. ARCHITECTURE: Data Layer (MongoDB) -> Ingestion Pipeline -> Validation & Drift Detection -> Transformation -> Training (GridSearchCV) -> MLflow Registry -> FastAPI Inference ENGINEERING DECISIONS: - Schema validation prevents silent data corruption across pipeline runs - KNN Imputer preserves feature relationships better than mean-fill baselines - MLflow artifacts support rollback, comparison, and reproducible experiments - FastAPI provides a production-friendly inference surface with OpenAPI docs PRODUCTION RELEVANCE: Modular, environment-reproducible, and API-deployable. Demonstrates the MLOps fundamentals expected from an early-career ML engineer.
LeafNetwork (Precision Agriculture)
PROBLEM: Agriculture teams need faster visibility into crop health, disease signals, and field-level risk, but most tools separate detection, monitoring, and recommendation workflows. SOLUTION: Designing a multi-agent agriculture intelligence platform that combines disease detection, satellite-aware monitoring concepts, and decision support for precision farming. ARCHITECTURE: Field Data -> Vision/Disease Detection -> Agent Orchestration -> Risk Summary -> Farmer Recommendations ENGINEERING DECISIONS: - Multi-agent design separates detection, monitoring, and advisory responsibilities - Vision models support plant disease identification from field imagery - Modular architecture keeps future satellite and weather signals easy to integrate PRODUCTION RELEVANCE: Shows applied AI product thinking across agriculture, remote sensing, and decision automation.
Tennis Analysis System with YOLO + PyTorch
PROBLEM: Sports footage contains player movement, ball tracking, and gameplay patterns that are difficult to analyze manually at scale. SOLUTION: Built a computer vision analysis system using YOLO and PyTorch concepts to detect key visual entities and support structured tennis performance analysis. ARCHITECTURE: Match Video -> Frame Extraction -> YOLO Detection -> Tracking/Feature Layer -> Analysis Output ENGINEERING DECISIONS: - YOLO supports fast object detection on match frames - Frame-level processing keeps the pipeline extensible for tracking and analytics - PyTorch-based workflow supports deeper model customization when needed PRODUCTION RELEVANCE: Demonstrates applied deep learning for sports analytics and real-time vision workflows.
Contact Me
Hiring for ML, AI engineering, or data science? Let's talk.
Contact Me
Open to machine learning, AI engineering, and data science internships where production mindset matters.