01 / INTELLECTUAL PROPERTY
01
Granted patent
Fuel gas cylinder wear and tear inspection system
AI ENGINEER / PRODUCTION SYSTEMS
Harsh
Agrawal.
I build and deploy AI systems that solve complex real-world challenges. I thrive on tackling hard problems, shipping production solutions, and delivering measurable impact.
CURRENT STATUSM.S. Urban Tech candidate · Program Representative at Cornell Tech
URBAN TECH / NYCFIELD NOTES 001
Data, models
& the built world.
& the built world.
02 / PROFESSIONAL EXPERIENCE
3+
Years of experience
Data science engineering, research, and applied machine learning
03 / CORNELL TECH
CORNELL
Merit scholarship
M.S. Urban Tech · Cornell Tech
MY JOURNEY / BEFORE CORNELL TECH
From idea to
systems that work.
I spent three years leading technical work at an AI startup, helping turn early concepts into deployable systems that solve real operational problems.
- ROLE
- Technical Lead
- DURATION
- 3 years
- TEAM
- Automation + Data Science
- IMPACT
- Top 20 · India AI Impact Summit 2026
01 / SELECTED PROJECTS
Applied work.
Machine learning, retrieval systems, and data for more resilient cities.
PROJECT / 01
FALL 2026
NYC FloodNet & Drainage Infrastructure Mapping
An interactive GIS project exploring urban flood vulnerability across Brooklyn Community District 6 (CD6), New York City, using FloodNet sensor data and drainage infrastructure datasets.
- Spatial data integration: Combined FloodNet sensor events and NYC catch-basin locations using NYC Open Data, ArcGIS Pro, and ArcGIS Online.
- Infrastructure analysis: Mapped 1,800+ catch basins and calculated density per square mile to examine drainage distribution.
- Event visualization: Mapped FloodNet event counts by sensor to identify locations with recurring recorded street-flooding events.
- Interactive StoryMap: Built a web map with thematic layers, density-based symbology, and a legend.
- Urban resilience: Compared event patterns with drainage distribution to identify areas for further investigation; this analysis does not establish that low catch-basin density causes flooding.
PROJECT / 02
FALL 2025 — SPRING 2026
Generative AI for Legal Document Retrieval & Chunking
A Generative AI framework for legal-document segmentation, context-aware retrieval, and grounded response generation.
- Intelligent chunking: Designed document segmentation approaches to preserve context across lengthy legal texts.
- Context-aware retrieval: Focused on improving the relevance of retrieved passages for RAG pipelines.
- Grounded generation: Enabled LLMs to generate responses using relevant legal-document context.
FEATURED PROJECT / INDUSTRIAL AI
Automated Cylinder Inspection.
Machine-learning quality inspection for LPG cylinders, built to improve detection and safety.
MAIN PROJECT / 01
GRANTED PATENT / LPG QUALITY INSPECTION
Automated LPG Cylinder Quality Inspection
Applied machine-learning pattern recognition to 3D point-cloud data to detect surface anomalies on LPG cylinders.
- Reached 91.7% accuracy in cylinder surface anomaly detection.
- Accelerated C++ inference with CUDA (950× speedup) and multithreading (1.5× faster) to enable real-time checks.
- Improved geometric search with KD-trees and queue-based data structures.
- Received a patent for “System and Method of Checking Fuel Gas Cylinders for Wear and Tear Damage.”
OTHER WORKS / 02 PROJECTS
Also built.
OTHER WORK / 01ROBOTICS / INDUSTRIAL AI
Robotics & Vision Inspection
Built robotic inspection workflows that integrated AI vision to detect defects across precision industrial components and production systems.
- Programmed robotic inspection workflows.
- Integrated AI vision for defect detection.
- Built quality checks for axle chassis components.
IMPACT
Improved product quality by rejecting defective units before downstream stages, helping the client avoid $2,000+ in daily penalties while reinforcing a safety-first inspection standard.
OTHER WORK / 02HVAC PREDICTIVE MAINTENANCE
Predictive Maintenance for HVAC
Built a predictive maintenance solution for a leading HVAC manufacturer using streaming sensor data.
- Developed predictive analytics models to identify maintenance needs from sensor streams.
- Used Kafka and Redis in the streaming data workflow.
- Reduced maintenance downtime and lowered maintenance costs by 23%.
IMPACT
23% lower maintenance costs, alongside reduced maintenance downtime.
03 / TECHNICAL PROFILE
Tools & systems.
Languages, platforms, and environments from my current toolkit.
01 / LANGUAGES
Python, C++, Java, R, HTML, CSS
02 / FRAMEWORKS & TOOLS
Kafka, Redis, MQTT, TensorFlow, ArcGIS, CUDA, AWS
03 / OPERATING SYSTEMS
Windows, UNIX, Linux
04 / EDUCATION
Education.
Cornell Tech, Cornell University
M.S. in Applied Information Science & Information Systems — Urban Tech
GPA 4.0 / Merit Scholarship
AI in Urban Data · Applied ML · Optimization Methods · Human-Computer Interaction
Vellore Institute of Technology
Bachelor of Computer Science Engineering
CGPA 8.62
Data Structures and Algorithms · Machine Learning · Data and Visualization
05 / CORNELL TECH LEADERSHIP
Community Leadership.
CORNELL TECH STUDENT GOVERNMENT / OCT 2026 — PRESENT
Program Representative, Urban Tech
- Represent the Urban Tech cohort in discussions and initiatives with Cornell Tech management.
- Organize cohort events and support initiatives that strengthen the Urban Tech community.
06 / RESEARCH PAPERS
Research.
07 / CERTIFICATION
Certifications.
AWS Certified Solutions Architect — Associate
JUL 2023 — JUL 2026
CONTACT / NEW YORK, NY