AI Sandbox.
A secure, governed environment to co-develop, test, and validate AI solutions for Haryana's public-sector challenges — proving what works before committing to scale.
What is the
AI Sandbox?
The AI Sandbox is a secure, high-performance environment designed to co-develop, test, and validate AI models for public-sector challenges. By pairing technology innovators with government departments and domain experts, it enables safe collaboration on real-world datasets.
Operating as a governed sandbox laboratory, it provides secure data access, specialized compute, and clear ethical guardrails for privacy and risk mitigation. Built on a "test before invest" philosophy, the program ensures only high-impact, verified solutions proceed to scale-up deployment.
Eligible Applicants
Tech Firms
Startups with active AI products.
Research Centers
Applied AI research groups.
Academic Labs
University public ML teams.
NGOs & Consortia
Collaborations for public benefit.
Excluded Categories
Early Stage
Pre-product or ideas without working ML.
Consultancies
Strategy or advisory agencies.
Non-AI Tech
Hardware or SaaS without core ML.

Review the implementation guidelines and vision document for the State AI Sandbox.
Five Priority Use Cases.
Each use case represents a critical public sector challenge with structured data resources, clear agency benchmarks, and a direct pathway to state-wide deployment.
"जींद में पानी की पाइपलाइन टूटी हुई है..."
Water Sanitation
98%
CMO Grievance Cell — Jansamvad AI
NLP for complaint comprehension, classification, and routing
NLP-powered system to comprehend, classify, and route citizen grievances filed through the Jansamvad platform — reducing resolution time and improving accountability.
Mhari Sadak + Urban Intelligence
Computer vision on citizen and satellite imagery for road defect detection
Computer vision applied to citizen-submitted and satellite imagery to detect road surface defects, prioritize maintenance, and monitor urban infrastructure health.
Teacher Deployment Optimization
AI-driven workforce planning across UDISE, assessment, and GIS data
AI-driven workforce planning that combines UDISE school data, learning assessment results, and GIS mapping to optimize teacher deployment across the state.
TB Eradication & Predictive Early Warning
Chest X-ray triage with HMIS / e-Upchar fusion for case finding and prediction
Chest X-ray AI triage integrated with HMIS and e-Upchar health records to improve TB case finding, enable predictive early warning, and support eradication goals.
Water Quality & Infrastructure Intelligence
Geospatial AI for monitoring, fault localization, and network mapping
Geospatial AI for water quality monitoring, infrastructure fault localization, and network mapping to improve service delivery by the Public Health Engineering Department.
Four Phases.
Go/No-Go Checkpoints.
The Sandbox acts as a structured validation funnel, guiding innovators from sourcing real needs to testing on real data, before deploying state-wide.
Discover
Collaborate with departments to audit data readiness, host workshops, and define key public-sector challenges.
Build
Launch the competitive call to select elite teams, granting them secure state data sandbox access and mentorship.
Validate
Rigorously test prototypes on simulated datasets while auditing for regulatory compliance and ethics.
Deploy & Scale
Transition validated models to live department pilots, monitoring real-world KPIs and scaling outcomes.
Program Milestones.
1 June 2026
Launch & Call for Proposals
Launched by the Hon'ble CM at Gurugram University
July 2026
Selected Teams Announced
Shortlisted innovators announced and onboarded
Oct–Nov 2026
Results & Demo
Results showcase and Sandbox Toolkit release
What You Receive.
Selected cohort participants gain access to specialized state resources, computing power, and regulatory scaling pathways.
Government Datasets
Curated, de-identified datasets for each use case — accessed within a secure sandbox environment under strict privacy protocols.
Cloud Credits & Compute
Compute resources and cloud infrastructure provided through technology partnerships. On-premise compute available for privacy-sensitive use cases.
Mentorship & Support
Domain experts, government counterparts, and AI mentors. Structured cohort sessions, peer reviews, and demo checkpoints.
Scale-Up Pathway
Solutions that meet defined KPIs during the pilot will have a clear pathway to scale-up deployment through government operations.
How Applications Are Evaluated.
Proposals are evaluated by an independent review board across six weighted criteria, optimizing for technical safety, deployment speed, and state impact.
Laser-focused use case alignment.
We prioritize solutions that exhibit a deep, structural understanding of target public services. Your team must demonstrate why a specific AI/ML approach is mathematically and operationally suited to solve the designated department's bottlenecks.
Our Selection Process.
From initial pre-screening to final board review, our selection timeline ensures absolute transparency and rigorous technical vetting.
Eligibility Pre-Screen
Applications screened against mandatory eligibility criteria. Ineligible applications are removed before scoring.
Independent Review
Each evaluator reads the full application before scoring. No partial scoring — holistic assessment first.
Evaluation & Scoring
Applications scored across six criteria on a 0–100 scale, with weighted totals calculated.
Selection Panel
Shortlisted teams reviewed by the jury. Selected teams announced and onboarded with data access and compute resources.
Application Form.
Complete the cohort submission. Selected teams receive secure data endpoints, GPU cloud credits, and operational sandboxes.
Integration Required
This form is a front-end preview. No submission endpoint or file storage is connected yet, so entries are not saved, transmitted, or received by anyone.
Pre-Screening
We verify eligibility first to ensure all cohort participants are formally registered in India, have applied AI/ML experience, and can commit a dedicated team to the 3-month sandbox cycle.