#338 Amith Singhee: Can India Catch Up in AI? IBM's Amith Singhee on What It Will Take

What It Covers

IBM Research India Director Amith Singhee examines why India has lagged in AI development despite abundant engineering talent, what conditions must converge for India to compete globally, and how IBM's enterprise-focused AI research — spanning hybrid cloud deployment, Granite LLMs, COBOL modernization, and agentic systems — addresses real-world business constraints.

Key Questions Answered

• •India's AI readiness formula: Three elements must converge simultaneously — sustained investment, deep tech talent, and consolidated GPU infrastructure — before India can compete in AI development. As of 2024, the India AI Mission has committed funding, but data center capacity remains in ramp-up, with consolidated clusters of 4,000-plus GPUs still being assembled and activated.
• •Enterprise AI deployment gap: Deploying frontier AI models inside regulated enterprises requires far more than model capability. Security, identity authorization, auditability, tool description quality, and on-premises portability all need engineering solutions. IBM's hybrid cloud architecture prioritizes model portability across any cloud or on-premises environment, giving regulated clients like banks control over where AI workloads run.
• •COBOL modernization as continual learning testbed: IBM's Watson Code Assistant for COBOL releases updated model versions every four to six weeks, making it a live production environment for continual learning research. The model explains legacy code in plain language, writes new COBOL, and translates COBOL to Java — addressing a critical risk as engineers fluent in legacy languages retire.
• •Low-data model customization techniques: When enterprise clients have limited proprietary data for fine-tuning, IBM researchers apply data mixing strategies, curriculum training sequences, and synthetic data generation to prevent catastrophic forgetting while embedding domain-specific knowledge. The goal is maintaining general model capability — keeping benchmark scores above 90 — while adding business-specific skills without full retraining from scratch.
• •Career strategy for the AI era: Engineers should prioritize domain fundamentals alongside AI fluency, since hiring managers consistently choose candidates who understand underlying principles and use AI over those who only use AI tools. Beyond skills, the ability to continuously learn and rapidly apply new knowledge is now the core career asset — a shift academic institutions need to structurally support.