NPCI - Data Science Role - Artificial Intelligence & Machine Learning - GenAI + Predictive (3-6 yrs)

Job Description: AI & Machine Learning (GenAI + Predictive)

What we do at NPCI ?

At the National Payments Corporation of India (NPCI), We have already solved for scale, powering the world's largest payment rail (UPI) with predictive models that make decisions in milliseconds.

We operate the world's largest real-time payment rails (UPI, processing 21.6 Billion+ transactions/month), backed by massive scale Predictive AI for fraud detection. We are also building our own LLMs and Fine-tuning other open-sourced LLMs for UPI Help and layering Industry grade Agentic AI on top of our infrastructure. Our goal is to move beyond simple transaction processing to create fully autonomous systems.

We need a visionary technologist who spans both worlds.

We are looking for a Principal AI Architect with 13+ years of experience to lead this convergence. You will not just build models, you will define the AI doctrine for the nation's financial backbone.

Why This Role? The Singularity of Payments

1. The Convergence of Disciplines:

Most roles ask for either "Classic ML" (Fraud/Risk) OR "GenAI" (LLMs). We need both. You will architect systems where Predictive Models (approving a transaction in 30ms) trigger Agentic Workflows (resolving a dispute autonomously via UPI Help). You will bridge the gap between deterministic security and probabilistic intelligence.

2. Unmatched Scale & Complexity:

You will oversee architecture for systems handling petabytes of data. Your challenge is to deploy 7B+ parameter LLMs alongside lightweight GBDT fraud models within the same millisecond-latency inference pipeline. This is an engineering challenge that exists nowhere else.

3. Nation-Scale Impact:

You will build the technology that brings banking to the "Next Billion Users" the non-English speaking population of Bharat. Your work on vernacular LLMs will democratize financial access.

Key Responsibilities: Strategy, Architecture & Execution

1. Strategic Leadership:

- AI Roadmap: Define the technical roadmap for integrating GenAI into core payment flows without compromising the 99.999% reliability of UPI.

- Build vs. Buy: Make high-stakes decisions on when to train proprietary foundation models vs. fine-tuning open-source models to ensure complete data sovereignty.

- Talent Magnet: Mentor Senior Engineers and Engineering Managers, fostering a culture of high-performance research and production excellence.

2. Predictive AI & Scale:

- Fraud at Scale: Architect the next generation of our Fraud Risk Management (FRM) system. Move beyond batch processing to Real-Time Graph Learning and streaming inference using Flink/Spark.

- Feature Store Architecture: Oversee the design of a unified Feature Store that serves both high-velocity transactional features and dense vector embeddings.

- Optimization: Drive kernel-level optimizations (CUDA, TRT-LLM) to squeeze maximum throughput from our GPU clusters.

3. Generative & Agentic AI:

- Autonomous Agents: Lead the development of Level 4 AI Agents capable of navigating banking interfaces, executing refunds, and detecting social engineering attacks in real-time.

- LLM Fine-Tuning Strategy: Direct the strategy for fine-tuning SLMs (Small Language Models) on secured financial data for edge deployment.

- Safety Guardrails: Design the "Constitution" for our AI immutable guardrails that prevent hallucination in critical financial advice.

The Ideal Candidate Profile

You are a "Builder-Leader." You have spent the last decade coding, architecting, and scaling, and you are still hands-on enough to review a PR or debug a distributed training crash.

Essential Experience

- Dual Competency : Proven track record in Classic ML (at scale) AND Modern GenAI (LLMs/Agents). You must have shipped both types of systems to production.

- Scale Veteran : You have architected systems handling 10,000+ TPS. You understand the fallacies of distributed computing.

- Deep Learning Depth : You don't just use libraries; you understand the math. You have likely trained custom architectures or heavily modified standard ones (Transformer variants, GNNs).

- Leadership : Experience leading cross-functional teams of 20+ data scientists and engineers. You know how to align research goals with business P&L.

Technical Stack Authority

ML Ops : Kubernetes, Kubeflow, Ray, MLflow.

Big Data : Spark, Kafka, Flink, Iceberg/Hudi.

GenAI : vLLM, LangGraph, DeepSeed, Unsloth.

Cloud/Infra : Deep understanding of GPU/TPU orchestration and cost optimization.

The Executive Package

Influence : Direct access to the C-suite to shape the organization's technological future.

Resources : Access to state-of-the-art compute infrastructure and proprietary datasets that are unique in the world.

Legacy : The opportunity to say, "I architected the AI that powers the Indian Economy."

Join us. Define the future of Intelligent Money.