The Skill India Assistant (SIA) is an AI-enabled multilingual chatbot developed to support users of India's national public skilling ecosystem. Launched on 17 May 2025 by the Ministry of Skill Development and Entrepreneurship (MSDE), SIA was developed through a partnership between Meta, SarvamAI (as the implementation partner), and the National Skill Development Corporation (NSDC), which operates under MSDE. SarvamAI, an Indian AI company specialising in multilingual language models, served as the technical implementation partner responsible for building and deploying the solution. The chatbot is built on Meta's open-source LLaMA (Large Language Model Meta AI) foundation model and is described as the first large-scale deployment of open-source AI within a national public skilling mission.
SIA is accessible through two primary channels: the Skill India Digital Hub (SIDH), which is the government's central digital platform for skilling services, and WhatsApp, the widely used messaging application. Users can access SIA on WhatsApp by contacting +91 8448684032. The decision to deploy SIA on WhatsApp is strategically significant, as WhatsApp has deep penetration across India including in rural and underserved areas where access to formal digital platforms may be limited. This dual-channel approach is intended to maximise reach and ensure that users who may not regularly access the SIDH web portal can still benefit from the AI assistant through a platform they already use daily.
The core functionality of SIA centres on four main features. First, the chatbot provides tailored course recommendations, helping users identify relevant skilling and training programmes based on their interests, qualifications, and career goals. Available courses span multiple domains including entrepreneurship, artificial intelligence, information technology, banking, and agriculture. Second, it includes a training centre locator that enables users to find nearby skill development centres, reducing information barriers that may prevent individuals from accessing training opportunities. Third, SIA offers job listings, connecting users with employment opportunities that match their skill profiles. Fourth, the system provides interactive quizzes that allow users to assess their knowledge and readiness in various skill domains. Together, these features position SIA as a matching and recommendation tool that connects individuals with appropriate skilling pathways and employment opportunities within India's labour market programmes.
SIA supports multilingual interaction in English, Hindi, and Hinglish (a hybrid of Hindi and English commonly used in everyday Indian communication), with both text and voice input modalities. This language support reflects the linguistic reality of India's workforce, where many users are more comfortable communicating in Hindi or in mixed-language registers rather than in English alone. The voice input capability further extends accessibility to users who may have limited literacy or who prefer verbal interaction. The chatbot operates on a 24/7 basis, providing continuous access to skilling information and recommendations without the constraints of office hours or staffing availability. The initiative is positioned within India's broader 'AI for All' vision and aligns with the country's digital public infrastructure strategy, which includes the Bhashini language platform for multilingual access.
From a technical perspective, SIA represents a foundation model deployment. The system is built on Meta's LLaMA, an open-source large language model, which has been adapted by Meta, SarvamAI, and NSDC for the specific requirements of India's skilling ecosystem. The use of an open-source foundation model is notable in the context of government AI deployments, as it provides a degree of transparency and adaptability that proprietary models may not offer. The model provenance is classified as adapted from open-source, reflecting the fact that while the underlying LLaMA model is openly available, it has been customised and fine-tuned for the Skill India context.
The development process involved a mix of in-house and third-party contributions. NSDC, as the government's implementing agency for skill development, provided domain expertise and programmatic oversight, while Meta contributed the foundation model and SarvamAI served as the implementation partner responsible for building and deploying the chatbot solution. This public-private partnership model is consistent with India's broader approach to leveraging private sector technology expertise for public service delivery.
The deployment of SIA sits within the implementation and delivery chain level of India's labour market programmes, specifically serving the functions of profiling, job matching and support services, as well as outreach, communications and sensitisation. By providing personalised course recommendations and job listings, SIA directly supports the matching of individuals to appropriate skilling and employment pathways. By making this information available through WhatsApp and in multiple languages, it also serves an outreach function, extending the reach of Skill India's services to populations that might otherwise face information barriers.
As a newly launched system, SIA's scale and impact data are not yet available from the source material. The deployment status is assessed as operational deployment with limited rollout, reflecting the fact that while the system is live and publicly accessible, comprehensive usage data and outcome metrics have not yet been reported. The decision criticality is assessed as low, as SIA functions as an informational and recommendation tool rather than a system that makes binding decisions about programme eligibility or benefit allocation. Users receive suggestions and information but retain full autonomy over their skilling and employment decisions.
From a risk perspective, the primary concerns relate to the inherent characteristics of large language model deployments. Hallucination or misinformation is a key risk, as the LLM could generate inaccurate course information, incorrect training centre details, or misleading job listings. Reliability across the supported languages (English, Hindi, Hinglish) requires ongoing validation to ensure consistent quality of recommendations. The system's dependence on Meta's LLaMA model and WhatsApp platform introduces upstream dependencies that could affect service continuity or create vendor-related constraints. Governance risks include the absence of publicly documented risk assessments, bias audits, or independent evaluations. No information has been identified from the source material regarding human oversight protocols, escalation pathways, grievance mechanisms, or data protection impact assessments for SIA.