Google DeepMind launched ATL Saathi on July 14, 2026: a web application powered by Google's Gemini AI that acts as a planning and training assistant for teachers running Atal Tinkering Labs in Indian government schools. The pilot starts with 100 schools.
Atal Tinkering Labs are dedicated workshop spaces inside schools, equipped with robotics kits, 3D printers, electronics, and coding tools. They are part of a government program called the Atal Innovation Mission, run by India's national planning body NITI Aayog. As of late 2025, there are 10,000 of these labs across India, engaging more than 11 million students and having produced over 1.6 million student innovation projects since the program launched in 2016.
The Indian government has since committed to a much larger rollout. In the 2025 budget, Finance Minister Nirmala Sitharaman announced funding to set up 50,000 new labs in government schools over the next five years, a five-fold increase in scope. The budget allocated to the Atal Innovation Mission for 2025-26 was approximately four times higher than the previous year. That pace of expansion is exactly where a tool like ATL Saathi becomes relevant: the lab count can grow, but trained, confident teachers cannot be conjured at the same speed.
The teacher gap in India's government schools is one of the most documented problems in the country's education system. A parliamentary committee reported roughly one million vacant teaching posts across government schools, with the shortfall concentrated in rural and remote areas. The teachers who are present are often managing multiple grades in a single classroom. Asking these educators to also lead sessions on robotics, IoT, and 3D printing without dedicated support is a real stretch.
ATL Saathi is designed to address that stretch directly. Teachers can use it to get summarized curriculum modules, AI-generated visual guides, and short video overviews, replacing the longer training materials that most cannot sit through. The tool generates ready-made project ideas aligned to a student's grade level. When a student brings a specific idea or problem, the teacher can ask ATL Saathi for step-by-step build instructions, wiring diagrams, and safety notes on the spot. The assistant works in 8 Indian languages at launch, responding in whichever language the teacher uses.
This is a teacher-facing tool, not a student-facing one. That distinction matters. Student-facing AI tutors carry more risk around accuracy, age-appropriateness, and misuse. A tool built to support the teacher, who still leads and filters everything, is a safer and often more effective starting point for large public education systems.
The broader Google DeepMind strategy behind this is clear. Since February 2026, when the company announced a formal national partnership with India's government, the focus has been on embedding AI into specific, high-leverage public programs rather than selling general products. The ATL Saathi pilot sits alongside other initiatives including AI research tools for Indian scientists and AI-assisted training for 20 million government employees.
For anyone running training, education, or knowledge-transfer operations in sectors outside tech, the ATL Saathi model is a useful reference. The core design: take an existing training curriculum, organize it inside an AI system, and give frontline practitioners a tool that answers their in-the-moment questions in their preferred language. That pattern is not specific to schools. It applies anywhere an organization has documented knowledge that is hard for practitioners to access quickly.
The 100-school pilot is small enough that the results will be interpretable. If teachers report spending less time preparing and more time actually teaching, and if student project quality improves, this will accelerate toward the broader ATL network. If the results are mixed, the tool will need iteration. Either way, watching what happens in those 100 schools over the next two quarters will give a clearer read on how much of this transfers.