Dr. Raman Khurana

Staff Scientist

Northwestern University

Evanston, IL  ·  USA

AI for Quantum  ·  Quantum for AI  ·  Time Series Forecasting  

AI for Quantum Quantum for AI Quantum Computing Generative AI / LLMs Agentic AI Time Series Forecasting Particle Physics
0+
Years Experience
0
Data Processed
0+
Int'l Conferences
0+
Students Mentored
Dr. Raman Khurana

Three Domains. One Vision.

Bridging experimental particle physics, modern AI systems, and quantum computing to solve frontier problems.

🤖

AI / ML / GenAI

Designing scalable GenAI systems, fine-tuning LLMs (LoRA/QLoRA), building RAG pipelines, and multi-agent architectures for Fortune 500 clients. Physics-inspired feature engineering for time series — submitted to NeurIPS 2026.

LangChainLangGraphRAG LoRA / QLoRAAutoGenCrewAI Time SeriesMCP / A2A
💡

AI for Quantum Computing

Neural networks for quantum state tomography (10× speedup), AI-driven polarization control for quantum fiber communication (95–99% fidelity), and benchmarking VQCs on NISQ hardware via gradient analysis.

QiskitPennylaneqBraid QSTVQCNISQ Transfer Learning
⚛️

HEP · CERN · Big Data

7 years on the CMS Experiment at CERN. Led Higgs boson searches, four dark matter analyses (monoHbb, monoH combination, MET scanning & reconstruction), hadronic tau reconstruction (HPS), and 10 PB pipelines.

Higgs BosonDark Matter Tau Reco (HPS)10 PB Pipelines BDTROOT / C++MET Reco

Physicist Turned AI Researcher

I am an AI Researcher (Staff Scientist) at Northwestern University's Department of Electrical & Computer Engineering, with a PhD in Experimental High Energy Physics and 18+ years of experience in machine learning, statistical modeling, and large-scale data analysis.

My work spans three interconnected frontiers: modern AI and GenAI systems — LLM fine-tuning, RAG, and multi-agent architectures for Fortune 500 clients; experimental particle physics — processing 10 petabytes of proton-proton collision data at CERN; and AI for quantum computing — neural approaches to quantum state reconstruction, signal correction, and hardware benchmarking.

Before Northwestern, I spent 7 years at CERN as a postdoctoral researcher. I led four dark matter searches and orchestrated the mono-Higgs combination across five global teams (20 senior scientists), producing the world's best upper limit on dark matter production cross-section.

My PhD contributed directly to the discovery of the Higgs boson. I developed a novel analysis strategy published in JHEP and was the first Indian PhD student to independently lead a CMS analysis from inception to publication at CERN.

I teach "Everything Starts with Data" (MLDS-400) in Northwestern's graduate ML & Data Science program and have mentored 6 PhD students and 20+ undergraduate and master's students.


Selected Work

Filter by domain. Spanning quantum ML, GenAI systems, and particle physics.

Quantum MLSubmitted · Frontiers in Optics

Automated Quantum State Tomography via Neural Networks & Transfer Learning

NN architecture trained on synthetic data with transfer learning to expedite QST in realistic lab environments — far faster than maximum likelihood estimation.

↑ 10× speedup over MLE baseline
Quantum MLSubmitted · Frontiers in Optics

AI-Driven Polarization Control for Quantum Fiber Communication

State-space model that tracks and corrects quantum signals using a classical reference + time series modeling on synthetic and real fiber data. No direct quantum measurement required.

95–99% fidelity correction
Quantum MLPreprint in Progress

Characterization of NISQ Hardware via Gradient Analysis of VQCs

Benchmarking superconducting backends for variational quantum circuits across devices and circuit depths. Quantifies gradient stability as a proxy for trainability on near-term hardware.

VQC learnability across multiple hardware platforms
AI / MLSubmitted · NeurIPS 2026

Physics-Inspired Automated Feature Engineering via Navier–Stokes Decomposition

Automated feature engineering that projects time series into a high-dimensional space using fluid-dynamics mathematics (NS decomposition) to capture complex temporal dynamics.

↑ 20% over TSFuse & Catch22 baselines
AI / ML · Time SeriesDraft in Preparation

Enhancing Transformer Forecasting with Static Covariates

Extended Autoformer, Fedformer, and DLinear to handle heterogeneous data (static + temporal covariates). Multiple integration strategies designed and evaluated. Insights shared with Allstate.

↑ 6% forecast accuracy
GenAI · AgenticDeployed

Agentic AI for Real-Time Marketing Data Analytics

Multi-agent system (AutoGen + open-source LLMs) that translates natural language into real-time analytics workflows. Agents handle ingestion, summarization, code generation, and visualization.

90%+ accuracy · 60–80% reduction in analytics response time
GenAI · NLP · SQLDeployed

Text-to-SQL Platform for Enterprise Airline Data

Led a team of 4 master's students. Fine-tuned Gemini for NL-to-SQL over enterprise databases with strategic cost and latency reduction.

30% → 85% query accuracy · 50–80% cost reduction
GenAI · RAGDeployed

Insurance Document Intelligence Assistant

RAG-based assistant with PII guardrails, toxicity filtering, hallucination detection (DeepEval, AgentEval), and LLM-as-judge evaluation for a Fortune 500 insurance client.

70–90% time savings · 85%+ accuracy
HEP · Dark MatterPublished · CMS

Mono-Higgs Dark Matter Combination (monoH)

Orchestrated combination of five mono-Higgs analyses across five global teams (20 senior scientists). Unified heterogeneous datasets and statistical models into a single framework.

World's best upper limit on DM production cross-section
HEP · Dark MatterPublished · CMS

Dark Matter: monoHbb · MET Scanning · MET Reconstruction

Led four DM search projects. ML-driven signal-background discrimination on highly imbalanced 10 PB datasets. Novel analysis category boosted exclusion limits by 20%.

↑ 20% exclusion limit · 4 publications
HEP · ReconstructionPublished · ICHEP 2012

Hadronic Tau Reconstruction — HPS Algorithm

Developed and optimized a multivariate algorithm (Hadron Plus Strip) for detecting hadronic tau decays in CMS proton-proton collision events. Presented at ICHEP 2012, Melbourne.

>60% accuracy · ↑ 20% efficiency over prior methods
HEP · PhD ThesisPublished · JHEP

Higgs Boson Search in Uncovered Phase Space

Novel analysis strategy for the Higgs boson search using multivariate methods. First Indian PhD to independently lead a CMS analysis from inception to publication.

Published in JHEP · 4 international conferences

Multi-Agent AI Data Scientist

Open-source LLMs autonomously driving a full data science pipeline — from ingestion to insights to visualization.


Career Timeline

Sept 2022 — Present Northwestern University

ML Researcher — Staff Scientist / Research Associate

Dept. of Electrical & Computer Engineering · Evanston, IL

  • Quantum ML research: automated QST (10× speedup), AI polarization control (95–99% fidelity), and NISQ benchmarking via gradient analysis.
  • Physics-inspired time series feature engineering via Navier–Stokes decomposition — submitted to NeurIPS 2026 (20% improvement over SOTA).
  • Deployed multi-agent GenAI systems for Fortune 500 clients achieving 60–90% time savings at 85–90% accuracy using LangGraph, AutoGen, and open-source LLMs.
  • Advising on GenAI adoption: LLM safety, RAG pipelines, hallucination detection, and evaluation frameworks.
  • Instructor for MLDS-400 "Everything Starts with Data" — curriculum designed for 55 students.
  • Awarded 500K GPU/CPU compute hours for quantum ML simulation.
March 2015 — Aug 2022 CERN · CMS Experiment

Postdoctoral Researcher

Florida State University, USA & National Central University, Taiwan

  • Led four dark matter searches (monoHbb, mono-H combination, MET scanning, MET reconstruction) to publication in international journals.
  • Orchestrated mono-Higgs combination across 5 global teams (20 senior scientists)world's best upper limit on DM production cross-section.
  • Engineered Python pipelines for 10 PB of proton-proton collision data; 10× processing speedup via NumPy vectorization.
  • BDT for jet discrimination (10% sensitivity gain), jet energy regression (10% resolution improvement), HPS tau algorithm (20% efficiency gain).
  • GEANT4 simulation for HGCAL detector — 5–20% energy resolution gain. Real-time outlier detection reducing spurious events by 10–25%.
  • Educator at CMS Data Analysis School, Fermilab (2020). Awarded Ramanujan Fellowship, $150K (SERB, 2021).
Aug 2009 — March 2015 CERN · CMS Experiment

PhD Researcher

Saha Institute of Nuclear Physics · University of Calcutta · CERN, Geneva

  • Novel analysis strategy for Higgs boson search in uncovered phase space — published in JHEP.
  • Contributed to the discovery of the Higgs boson (2012) as a CMS collaboration member.
  • First Indian PhD to independently lead a CMS analysis from inception to international publication.
  • Awarded 18 months onsite at CERN across 6 three-month extensions for extraordinary performance.

Stack & Tooling

GenAI Stack
LangChainOllama OpenAI APITransformers HuggingFacePEFT LoRAQLoRA
Agentic AI
LangGraphAutoGen PhidataCrewAI MCPA2A
Quantum Computing
QiskitPennylaneqBraid
Machine Learning / AI
PyTorchScikit-learn XGBoostNumPy PandasStatsModelsNLTK
Big Data
PySparkPyFlink KafkaPostgreSQL CassandraMinIO
Languages
PythonC++ CSQL
Deployment
DockerKubernetes AWSCI/CDGitHub
HEP & Visualization
ROOTGEANT4 MatplotlibSeaborn TableauGrafanaLaTeX

Awards & Fellowships

🏆

Breakthrough Prize in Fundamental Science

2025 · "The Oscars of Science"

Awarded for unprecedented contributions to the discovery and precision measurement of Higgs boson properties as a member of the CMS collaboration at CERN.

🔬

Ramanujan Fellowship

2021 · SERB · $150K Research Grant

Prestigious national fellowship to carry out a dark matter search program and CMS detector upgrade contributions.

💻

500K GPU/CPU Compute Hours

Northwestern University

Awarded for quantum machine learning simulation and development of new ML models for quantum communication research.

⚖️

Judge — DataFest 2025

2025 · American Statistical Association

Selected as a national judge for the DataFest data science competition by the American Statistical Association.

📋

Peer Reviewer

Ongoing

Active referee for the Journal of High Energy Physics (JHEP), Data Science Journal (DSJ), and ICLR (session: AI4DiffEqtnsInSci).

🇨🇭

18 Months Onsite at CERN

2009–2015 · Geneva, Switzerland

Granted six consecutive 3-month extensions at CERN for extraordinary scientific performance — a rare distinction during PhD.


Selected Publications

For the complete list, visit Google Scholar.

2026

Automated Time Series Feature Generation via Profiling Frame and Navier–Stokes Decomposition

NeurIPS 2026 (submitted)
AI / MLSubmitted
2025

Automated Quantum State Tomography via Neural Networks and Transfer Learning

Frontiers in Optics (submitted)
Quantum MLSubmitted
2025

AI-Driven Polarization Control for Quantum Fiber Communication Networks

Frontiers in Optics (submitted)
Quantum MLSubmitted
2022

Search for new physics in events with a Higgs boson and missing transverse momentum — mono-Higgs combination and monoHbb channel

CMS Collaboration · Physical Review D / JHEP
HEP · CMSPublished
2015

Development of a Novel Analysis Strategy to Search for Higgs Boson in an Uncovered Phase Space

Journal of High Energy Physics (JHEP) · CMS Collaboration
HEP · PhDPublished
2012

Performance of the CMS Hadronic Tau Reconstruction (HPS) Algorithm in Proton-Proton Collisions

ICHEP 2012, Melbourne · CMS Collaboration
HEP · RecoPublished

Education Impact

Instructor — "Everything Starts with Data" (MLDS-400)

Northwestern University · Master in Machine Learning & Data Science · In Person

Designed and delivered full curriculum for 55 students with diverse academic backgrounds. Topics: data categorization, EDA, the ML lifecycle, feature engineering, tree models, SVMs, clustering, and optimization — with hands-on real-world projects.

Educator — CMS Data Analysis School

Fermilab, IL, USA · 2020

Taught best practices in data science and machine learning to undergraduate and PhD students at the CMS flagship analysis school at Fermilab.

Research Mentorship

Northwestern University & CERN · 2015 – Present

Mentored and supervised 6 PhD students and 20+ undergraduate and master's students on projects spanning wildfire prediction, time series forecasting, and agentic AI systems.


Academic Background

PhD — Experimental High Energy Physics

Saha Institute of Nuclear Physics, Kolkata · University of Calcutta · CERN, Geneva  |  Aug 2009 – March 2015

Dissertation: Development of a Novel Analysis Strategy to Search for Higgs Boson in an Uncovered Phase Space. Contributed to the discovery of the Higgs boson. First Indian PhD to independently lead a CMS analysis from inception to publication (JHEP). Awarded 18 months onsite at CERN. Presented at four international conferences.

MSc — Physics (Electronics)

University of Delhi

BSc — Physics (Honours)

University of Delhi

Get in Touch

Open to research collaborations, consulting engagements, and speaking opportunities.

📧

Email

ramankhurana1986@gmail.com
💼

LinkedIn

dr-raman-khurana
🐙

GitHub

github.com/ramankhurana
📖

Google Scholar

Publication record