Astrophysicist · Data Scientist · AI Engineer · Madrid

I turn data
into models, answers
& decisions.

PhD astrophysicist with 7+ years turning massive, noisy datasets into reliable results, from 100,000-source sky surveys to JWST spectroscopy. Today I build machine-learning and LLM systems end to end: data pipelines, models, evaluation, and automated deployment.

Nischal Acharya
Nischal Acharya, PhD
View my work → ✉ Request my CV
0+
Years with data
0k+
Sources modelled
0
Peer-reviewed papers
A&A and ApJ, one as first author
0
Languages spoken
Nepali, English, Spanish, German A1
01

Wherever there's data,
I find the signal.

I'm an astrophysicist based in Madrid and a former Marie Skłodowska-Curie Fellow. My research used large sky surveys and JWST spectroscopy to study active galactic nuclei: terabyte-scale data, statistical modelling, and reproducible Python pipelines running on high-performance computing clusters.

Today I bring the same rigour to machine learning and AI engineering. I've built an autonomous AI agent that picks the news, then writes, designs, narrates and publishes a daily news page with no human in the loop; a retrieval-augmented Q&A system with its own evaluation set; and ML emulators that predict the output of slow physical simulations.

Whatever the data, whether it's satellite imagery, market ticks or 100,000 galaxies, I care about the same things: reproducibility, honest evaluation, and results that people can act on.

Native Nepali speaker, fluent in English and Spanish, A1 German, and currently building Italian and Portuguese on Duolingo. Picking up new languages is a habit I enjoy.

02

Where I've worked.

Apr 2025 - Jul 2026
Postdoctoral Researcher
CEFCA, Teruel
  • Searched for low-luminosity active galactic nuclei in the J-PAS survey, working with multi-wavelength data on 100,000+ sources.
  • Built reproducible Python pipelines for ingestion, cleaning, and statistical modelling of survey photometry.
  • Measured emission-line fluxes from JWST observations and used their ratios to separate weakly accreting nuclei from Seyfert and star-forming populations.
  • Presented results to specialist and non-specialist audiences at international meetings.
Jun 2025 - Nov 2025
Quantitative Developer & Researcher
QuantumScaleMunich GmbH · Freelance, remote
  • Designed and back-tested trading strategies on crypto and equities, evaluating win rate, profit factor, and drawdown across thousands of trades.
  • Built the supporting Python stack: data-feed handlers, a strategy engine, and equity-curve and Sharpe analytics.
Sep 2020 - Feb 2025
PhD Researcher, Astrophysics (Marie Skłodowska-Curie Fellow)
Donostia International Physics Center (DIPC), San Sebastián
  • Doctoral thesis connecting AGN to the properties of their host galaxies with the miniJPAS survey, funded by the EU Horizon 2020 MSCA programme.
  • Designed a probabilistic selection pipeline that filtered 64,000+ sources into a high-purity sample of faint X-ray AGN, published as first author in Astronomy & Astrophysics.
  • Image processing, model fitting, and MCMC sampling on survey-scale data; talks at conferences and institutes across Europe.
Nov 2023 - Jan 2024
Research Data Analyst (Secondment)
Ludwig-Maximilians-Universität München
  • Built Python pipelines to extract, transform, and statistically characterise structural properties from the Magneticum cosmological simulations, with results used by a 10+ person research group.
Jul - Sep 2022
Data Scientist (Secondment)
OmegaLambdaTec GmbH, Garching
  • Trained in machine-learning classification of crop and land-cover types from European Sentinel satellite imagery.
  • Processed the multi-band rasters in Python (rasterio, pandas) and characterised the classified regions statistically.
Apr - Jun 2022
Visiting Researcher
Max Planck Institute for Extraterrestrial Physics, Garching
  • Multi-wavelength counterpart identification and classification of X-ray-selected galaxies in the high-energy group.
03

Things I've built.

🤖 Live

CryptoAssetsDaily: Autonomous AI News Agent

An AI agent that runs an Instagram news page with no human in the loop. Every four hours it ranks the latest crypto stories, writes a 4-slide carousel with an LLM, generates the artwork, and publishes it. Three times a day it also turns a story into a narrated Reel with synthesised voice, word-synced captions and music. It runs on scheduled GitHub Actions, entirely on free tiers, and tracks each Reel's voice and design variant for A/B analysis.

PythonGeminiFLUXText-to-SpeechWhisperffmpegGitHub ActionsInstagram Graph API
See it live →
📚

RAG Pipeline: Document Q&A

Retrieval-augmented generation over PDFs and text: chunking, embeddings, vector search and answers with cited sources, plus a benchmark that measures retrieval recall and answer quality.

LangChainChromaDBsentence-transformersLLMs
Code →
🧪

AstroLineRatios: ML Simulation Emulator

Gradient-boosted models trained on Cloudy photoionisation grids that predict JWST emission-line ratios for any physical conditions, without rerunning the simulations (validation R² 0.93).

scikit-learnGradient BoostingPython
Code →
🔭

JWST IFU Pipeline

Turns James Webb Space Telescope NIRSpec and MIRI integral-field data cubes into science-ready emission-line maps and diagnostics.

PythonAstropySpectroscopy
Code →
🧬

palfitology: Python Package

A tested Python package and CLI for PSF-aware isophote fitting across 12 J-PLUS bands, with per-object diagnostics, mosaics and GALFIT-ready outputs.

PythonAstropyphotutilspytest
Architecture → Code →
🌌

Faint X-ray AGN in miniJPAS

A probabilistic selection pipeline that filtered 64,000+ survey sources into a high-purity sample of faint X-ray AGN. First-author paper in Astronomy & Astrophysics.

PythonMCMCSED Fitting
Paper →
📈

Spikewatch: Real-Time Market Data Platform

A Python app for real-time monitoring of crypto markets: multi-exchange (Binance, Bitget) data ingestion, strategy back-testing with parameter sweeps, and Sharpe and drawdown reporting.

PythonQtWebSocketREST
Early access →
⚔️

sombra: Browser Action Game

A 2.5D hack-and-slash in the browser where every 3D model, sound and particle is generated in code, with no build step and no asset files.

JavaScriptthree.jsWebGL
Play → Code →
WortSwipe icon

WortSwipe: German Flashcard App

A swipe-based Android app for learning German vocabulary, in development. Built end to end with Claude Code, from design and architecture to product decisions.

KotlinAndroidAI-assisted development
🧩

CIGALE: Open-Source Module

Co-authored a paper introducing a reusable module for probabilistic SED modelling in the open-source CIGALE code.

Open SourcePythonPackaging
04

Peer-reviewed research.

A&A The miniJPAS Survey: The radial distribution of star formation rates in faint X-ray active galactic nuclei First author · Astronomy & Astrophysics · 2024 ↗ ApJ ReveaLLAGN 1: JWST emission-line spectra reveal low-luminosity AGN with ultraviolet-deficient SEDs and warm molecular gas Co-author · The Astrophysical Journal · 2026 ↗ ApJ Galaxy and Mass Assembly (GAMA): The properties of quasar host galaxies, star formation histories and stellar populations Co-author · The Astrophysical Journal · 2026 ↗ A&A The miniJPAS survey: AGN and host galaxy co-evolution of X-ray selected sources Co-author · Astronomy & Astrophysics · 2023 ↗ ORCID Full publication record, including GAMA quasar-environment papers in ApJ Complete list on ORCID ↗
05

Trained to think rigorously.

2020 - 2025
PhD, Astronomy & Astrophysics
University of the Basque Country (EHU), research at DIPC, San Sebastián
Marie Skłodowska-Curie Fellow. Thesis connecting AGN to the properties of their host galaxies: large-scale survey data, statistical modelling, reproducible Python pipelines. Published in A&A and ApJ.
2017 - 2020
MSc, Astronomy & Space Physics
University of Turku, Finland
Thesis: statistical analysis of 200,000+ galaxies: feature engineering, hypothesis testing, MCMC.
2011 - 2015
BSc, Physics
St. Xavier's College, Kathmandu
Key coursework: Statistics I & II, Integral Calculus & Differential Equations, Real Analysis, Advanced Algebra & Matrices, Quantum & Mathematical Physics, Classical Mechanics, Electronics, and Research Methodology.

Let's build something
with your data.

The fastest way to reach me is email. I reply quickly.