C:\My Documents\CV.htm Full Professional Log ↗

Luis Ángel Quiroz López

Data Science · Data Engineering · AI

Summary

Physicist with an MSc in astrophysics (UNAM) who builds data and AI solutions in Python and SQL, from data pipelines and dashboards to LLM applications with RAG. Freelance experience evaluating LLM outputs against quality rubrics for model training (RLHF). Experienced in presenting technical work to non-specialist audiences.

Selected projects

Reward hacking on GSM8K: two small model organisms

Python · PyTorch · Transformers · PEFT (LoRA) · Colab

SFT, DPO and GRPO written from scratch in PyTorch with Claude Code, then used to study what a 0.5B language model learns when its RL grader has a loophole: one grader also pays the answer the user suggested, the other pays any final line that contains the right number.

  • A grader that also pays the answer the user suggested drove copying of wrong suggestions from 7% to 97% of training samples. The habit carried over to unseen phrasings (64%) and to SVAMP problems (62%).
  • In 89% of the copies the reasoning itself arrives at the user's number, usually through a false equation, which a consistency monitor cannot see. Penalising what that monitor flags cut flagged copies from 10.8% to 1.8% while copying stayed at 96%.

Incentive compensation API: a QA case study

Postman · Newman · pytest · PostgreSQL · FastAPI · GitHub Actions

A full QA cycle on a sales-commission REST API, the kind of calculation that sales performance management software automates: split credit, clawbacks on returns, a 50% gate, tiered accelerators and a cap.

  • 92 risk-based test cases traced to the requirements, and 8 defects found in the first build, 2 of them critical.
  • A parallel run against an independent Excel model measured 6,566.74 MXN overpaid by the first build and 750 of 750 payouts correct after the fixes.

Vigía: operational stability monitoring

Java · Spring Boot · Flask · Django · PostgreSQL · MongoDB · Docker

A small polyglot platform that ingests request telemetry, reports service-level objectives and detects anomalies: Spring Boot ingestion, a Flask analytics API, a Django back office, PostgreSQL and MongoDB.

  • 1.4 million simulated requests over ten days with three injected incidents, all three detected and nothing else.
  • A Monte Carlo calibration over 40 runs cut false alarms from 18 to 1 without missing an incident.

Lecture notes pipeline with agentic RAG

Python · faster-whisper · Claude Code · Anthropic API · ChromaDB · ffmpeg

Turns four-hour class recordings into structured Obsidian notes with no manual steps, then answers questions about past classes citing the class and the minute: local transcription on the CPU, notes written by Claude Code in headless mode and an agent over hybrid retrieval.

  • Fits in 8 GB of RAM: silence-aligned ten-minute chunks cut peak audio memory from more than 6 GB to about 0.7 GB on a two-hour recording.
  • Hybrid retrieval over notes and transcripts, with multilingual embeddings and Spanish BM25 merged by Reciprocal Rank Fusion. The agent answers only with real citations or says the notes do not cover it.

Retail delivery diagnostics: an end-to-end case

Python · pandas · SQL · Star schema · Chart.js · Statistics

Diagnosis of a department-store delivery network, from 62,420 raw shipment records to an executive dashboard. The monthly report said 95.7% of deliveries were on time and in full, yet management felt deliveries were getting slower. The analysis shows where and why.

  • Cleaning with lineage: 62,420 raw records reduced to 59,615 shipments, with every exclusion logged.
  • Stores compared with their own zone through Wilson intervals, so only 1 of 61 stores is flagged as significantly worse.

EMA/MACD signal toolkit

Python · pandas · Excel · pytest · Bitso and Binance APIs

Moving averages, MACD(12, 26, 9) and RSI implemented from first principles for a finance diploma, then extended into a batch run over 35 equities listed on the Mexican Stock Exchange and a bitcoin cost comparison across Mexican platforms.

  • Cross-validated against a 3,815-formula Excel model to within 5 × 10⁻¹³, recalculated in CI on every push.
  • Bitcoin with pesos: live Bitso and Binance order books against Mercado Pago, measured from the prices its app shows. Buying and selling at once costs 0.27% on Binance and 3.98% on Mercado Pago.

Small language model from scratch

Python · PyTorch · NumPy

Every component of a small language model, implemented and then trained, for Carnegie Mellon University's Introduction to Modern AI (online, January to April 2026).

  • An autodiff engine, a BPE tokenizer and the Adam optimiser written from first principles.
  • A decoder-only transformer with masked multi-head attention, a KV cache, RMSNorm and a gated MLP, trained on TinyStories.

Milky Way structure: Gaia DR3 × APOGEE-2 DR17

Python · pandas · astropy · scikit-learn · TOPCAT

Spatial and kinematic characterisation of the Milky Way's stellar populations, built by cross-matching Gaia DR3 astrometry with APOGEE-2 DR17 spectroscopy.

  • 682,284 sources reduced to 281,612 stars through documented cuts on signal-to-noise, astrometric fit, radial-velocity scatter and parallax.
  • Velocity ellipsoid tilted by a vertex deviation of about −26.6°, consistent with asymmetric drift.

Galaxy morphology with convolutional networks

Python · TensorFlow / Keras · scikit-learn · Transfer learning

Binary classification of 5,608 MaNGA galaxy images into early types and late types, from a baseline CNN to transfer learning, with the class imbalance handled explicitly.

  • Baseline CNN: 81.0% test accuracy, held back by a recall of only 0.68 on late types.
  • A deeper CNN reached 86.3% test accuracy and raised late-type recall to 0.77.

Core skills

Programming
Python (pandas, NumPy), SQL, R, Bash, JavaScript, HTML/CSS
Data engineering & cloud
PostgreSQL, data modelling (star schema), ETL, data validation, Spark, Snowflake, Azure ML, FastAPI, Flask, Git, GitHub Actions, Linux
Machine learning
scikit-learn, PyTorch, TensorFlow/Keras, statistical inference (maximum likelihood, model selection, hypothesis testing)
LLMs & RAG
Anthropic API (streaming, tool use, token counting), Claude Code (headless mode, skills), prompt engineering, RAG with ChromaDB and hybrid retrieval, LLM evaluation (rubrics, red-teaming), RLHF, SFT and DPO
Quality assurance
Test design (boundary values, equivalence partitioning, negative testing), API testing with Postman and Newman, pytest, SQL data validation, regression testing in CI, Jira
BI & productivity
Chart.js, Excel (pivot tables, data validation), Power BI and Looker Studio (basic), Google Workspace, Microsoft 365

Experience

AI Trainer, Software Validation & NLP2026
DataAnnotation · Remote

Tested unreleased LLMs across 10+ projects with adversarial prompts that exposed logic failures and policy violations, reporting each one in expected-versus-actual form. Designed the rubrics and metrics used to score model outputs, and provided RLHF feedback and NLP annotations for model evaluation and fine-tuning.

Language Specialist and Freelance AI Trainer2026
Meridial Marketplace · Remote

Reviewed and corrected LLM outputs in multilingual human-in-the-loop pipelines, giving structured feedback used to guide training.

Freelance Field Data Verification Specialist2026
Independent contract · Mexico City

Collected and verified data on commercial venues through site visits and digital documentation, keeping the dataset accurate and complete.

Freelance Web Developer2025
Ashland Chemical de México · Mexico City

Built an interactive product-presentation website for private-school outreach, from client requirements to a deployed interface.

Research Assistant2024
UNAM, with Dr Gabriella Piccinelli Bocchi and Dr Ángel Sánchez Cecilio · Mexico City

Reviewed the cosmology literature, wrote a literature review on cosmological perturbations and presented the theory to the research group every week.

Science Communicator (volunteer)
Institute of Physics, UNAM · Mexico City

Presented cosmology research to audiences ranging from children to researchers in other fields, and built a live-camera demo that simulates a gravitational wave passing through the viewer.

Education

MSc in Field and Particle Astrophysics2026
Institute of Astronomy & Institute of Physics, UNAM

Thesis: Effective Equation of State and Gravitational Particle Production in Inflationary Models, supervised by Dr Marcos A. García García.

BSc in PhysicsAug 2019 to Dec 2023
National Autonomous University of Mexico (UNAM)

Degree awarded for high academic performance (titulación por alto desempeño académico), with an honourable mention (mención honorífica).

Courses & certifications

  • Introduction to Modern AI, Carnegie Mellon University (online), January to April 2026
  • Diploma in Finance, UNAM, March 2026 to present
  • zero2hero (2026): Claude Essentials · Claude Architect · Project Management and AI Agents with Claude · Software Development with Python and AI (24 sessions) · FastAPI and SQLModel · Financial Algorithms with Python · Statistics, Time Series and Econometrics with R

Languages

Spanish (native) · English (C2) · Italian (B2) · German (B1) · Portuguese (B1)