Applied Research Director - Advanced Algorithms and Compute

Niraj Kumar

I lead the Advanced Algorithms & Compute team in JPMorganChase Global Technology Applied Research, with research spanning advanced algorithms for Data-anayltics, Optimization & Pricing problems, Agentic-AI & Open-weight AI model research, and AI accelerators (GPUs, custom ASICs).

View publications Google Scholar LinkedIn
JPMorganChase Head of Advanced Algorithms & Compute.
Past Companies Quantum Algorithms Developer at PayPal and Pasqal.
PhD Quantum Computing, Telecom ParisTech.
Undergrad Bachelor's & Master's in Physics, Indian Institute of Technology Kanpur.

Research Focus

My current research spans advanced algorithms, AI compute platforms, and quantum computing, with an emphasis on methods that are useful in practical systems.

Advanced Algorithms

Developing advanced algorithms for applications in data analytics, optimization, pricing, and machine learning.

  • Data analytics
  • Optimization
  • Pricing
  • Machine learning

Semiconductors & AI

Work on agentic AI and open-weight models with GPUs and other emerging AI accelerators, with attention to model capability and compute efficiency.

  • Agentic AI
  • Open-weight models
  • GPUs
  • AI accelerators

Quantum Computing

Research across quantum algorithms, QAOA, privacy in distributed learning, quantum communication, cryptography, and quantum computing experiments.

  • Quantum ML
  • QAOA
  • Cryptography

Recent Papers

Selected recent papers related to my current research areas. See Google Scholar for the full publication list.

CRUMB: Efficient Prior Fitted Network Inference via Distributionally Matched Context Batching

Architecture-agnostic context batching for scalable tabular foundation model inference.

Read paper

Entropy Distribution as a Fingerprint for Hallucinations in Generative Models

Single-pass, black-box hallucination detection using calibrated token-level entropy distributions.

Read paper

Anytime Training with Schedule-Free Spectral Optimization

Schedule-free spectral optimization for high-quality checkpoints without committing to a fixed training horizon.

Read paper

Fast provable clustering for big-data analytics

Quantum-inspired classical methods for scalable clustering and analytics.

Read paper

Tensor-based methods for fine-tuning Large Language Models

Novel MetaTT method for parameter efficient fine-tuning of LLMs.

Read paper

Certified quantum randomness generation

Experimental work on certified randomness generation using quantum hardware.

Read paper
Explore publication highlights Full Scholar profile

Journey

Professional experience across quantum information, startups, payments, and applied research.

  1. Applied Research Director, JPMorganChase

    Executive Director in Global Technology Applied Research, leading Advanced Algorithms & Compute.

  2. Applied Research Lead, JPMorganChase

    Vice President, working across quantum algorithms and applied machine learning research.

  3. Senior Quantum Algorithms Researcher, Pasqal

    Developed algorithms for neutral-atom quantum computing applications.

  4. Quantum Algorithms Researcher, PayPal

    Explored algorithmic approaches for quantum and classical financial technology use cases.

  5. Postdoctoral Researcher, University of Edinburgh

    Research in quantum computing, communication complexity, and quantum information.

  6. PhD, Telecom ParisTech

    Design, analysis, and implementation of advanced quantum communication protocols.

  7. Bachelor's and Master's in Physics, IIT Kanpur

    Physics training with early hands-on work in robotics and systems.

Research in the Press

Selected coverage of research projects in quantum advantage, randomness, and communication.

Wall Street Journal Coverage of a quantum streaming algorithm demonstration using Quantinuum ion-trap hardware. Bloomberg Coverage of certified quantum randomness generation reported in Nature. Quanta Magazine Quantum communication advantage with photonic systems. Wired Experimental quantum communication advantage and its implications.