Researcher · Astrophysics & Scientific Computing

Dr. Arman Khalatyan

Leibniz Institute for Astrophysics Potsdam (AIP)

Building the data infrastructure for precision astrophysics — from 400 million-star catalogues to the HPC, ML and AI systems that fit, benchmark and ship them.

Portrait of Dr. Arman Khalatyan

01 On this site

01 Publications 34 works — from compact stars to 400-million-star Milky Way mapping. 895+ citations, h-index 11. → 02 Projects PhysicsLLM, SH26, PMViewer, VR@AIP, DRP-Hub (PUNCH4NFDI) and more — research in depth. → 03 Benchmarks S3 object-storage and LLM/SQL-generation benchmarks with full result plots. → 04 Gallery Color–magnitude diagram of 217 million SHBoost-labeled Gaia stars (2024 press release). → 05 Software Pipelines, viewers, agents and HPC utilities — 13 repositories. → 06 Blog Essays on astrophysics, AI and research — latest: Eight GPUs, Two Generations: Benchmarking Open LLMs for an Astronomy Institute → 07 Contact Email, GitHub, ORCID, LinkedIn, institute page. →

02 Research & Work

01 Stellar Parameters & Extinction at Scale

Full stellar-parameter and dust-extinction posteriors for hundreds of millions of stars (StarHorse / SHBoost), chrono-chemical mapping of the Milky Way disc, bulge and bar with Gaia DR3, APOGEE and 4MOST, and the 4MIDABLE survey design for 4MOST.

  • StarHorse
  • Gaia
  • 4MOST
  • Milky Way
02 Galaxy Formation & the Local Group

High-resolution cosmological simulations of the Local Group and early galaxy assembly — from the first thin discs at z > 4 to AGN feedback in galaxy evolution (MareNostrum, Hestia). Earlier work on the QCD phase diagram and compact stars.

  • N-body
  • simulation
  • AGN

03 Technical Expertise

HPC & Cluster Computing

400+ node SLURM clusters, fault-tolerant large-scale jobs, NUMA/RDMA, throughput optimization.

Data Engineering

PyArrow, Dask, Parquet, S3/MinIO, 210 GB full-catalogue pipelines, end-to-end checksum verification.

Machine Learning & MLOps

vLLM inference, LoRA fine-tuning, model quantization, evaluation harnesses, agentic orchestration.

Astrophysics Research

Stellar parameter estimation, extinction mapping, galaxy formation, N-body simulation, spectroscopic surveys.

DevOps & Systems

Docker, Kubernetes, CI/CD, HAProxy, monitoring, Proxmox, secure multi-host infrastructure.

Languages & Tooling

Python (primary), Bash, C/C++ (simulation, OpenGL), LaTeX, Git, Jupyter, Matplotlib.

04 Recent Publications

The five most recent — the full record lives on the publications page.

  • 2026 The Nineteenth Data Release of the Sloan Digital Sky Survey DOI ↗
  • 2026 The spheroidal bulge of the Milky Way: Chemodynamically distinct from the inner-thick disc and bar DOI ↗
  • 2025 Latest Developments of the PUNCH4NFDI Compute and Storage Infrastructures DOI ↗
  • 2024 Discovery of the local counterpart of disc galaxies at z > 4: The oldest thin disc of the Milky Way using Gaia-RVS DOI ↗
  • 2024 Insights from super-metal-rich stars: Is the Milky Way bar young? DOI ↗

05 Contact

Email … GitHub github.com/arm2arm → ORCID 0000-0002-8913-0690 → LinkedIn linkedin.com/in/arm2arm → Institute AIP — Leibniz-Institut für Astrophysik Potsdam →