While most teenagers spend Class 12 worrying about exams and college admissions, a 19-year-old from Bihar claims he spent the last two-and-a-half years building a massive artificial intelligence model from scratch using family savings, compute grants and countless failed experiments.
Abhinav Anand says in a social media post that he has developed a 5.82-billion-parameter multimodal AI model called “ArcleIntelligence” regardless of having no formal computer science degree, no investors and no team backing him.
How did he build the model?
Anand said that just a few years ago, his understanding of AI was limited to hearing about ChatGPT. Over time, he began experimenting with different projects including a YouTube analytics app, a voice assistant and an offline AI assistant before moving to larger AI systems. He explained the journey as a long process filled with failed attempts that eventually helped him learn how AI systems work in real-world conditions.
Anand’s ArcleIntelligence
According to Anand, “ArcleIntelligence” is designed as a multimodal model capable of processing text, images, documents, audio and video. He claimed the system can generate images at 512×512 resolution, produce speech output at 24kHz and support a context window of over 2 million tokens.
Anand also claimed that the model scored 93.45 on OmniDocBench V1.5 during private testing. However, the benchmark performance has not yet been independently verified. Before starting work on the multimodal model, Anand reportedly trained a text-to-video AI system on his laptop without external funding and later released it publicly through Lightning AI as a studio template.
According to Anand, the project was funded through personal savings, RunPod compute grants, DigitalOcean credits and GitHub Student Pack benefits. He estimated that GPU compute expenses alone cost nearly Rs 64,000, which he described as a major financial burden for his middle-class family in Bihar. Anand said his father works as a government officer while his mother is a homemaker.
Looking for more funds
Anand said the model is still being trained and that he is currently seeking around $35,000 in funding to complete the AI pipeline. He added that he plans to release the model weights on Hugging Face and eventually make the full codebase open source on GitHub.
Disclaimer: The content in this article is based on a viral social media discussion and is intended for informational and entertainment purposes only. The financial figures and strategies mentioned are personal to the user and have not been independently verified. This story does not constitute financial advice or an endorsement of any specific investment strategy. Readers are advised to consult a SEBI-registered investment advisor before making financial decisions.