Nick Ashouri

AI & LLM Engineer · BI Developer

AI/LLM engineer currently shipping an OpenAI + LangChain pipeline on Azure that handles about 100 medical documents a day for a German insurance-rehabilitation service — ~90% average accuracy, 30 minutes of case-handling compressed to seconds. A parallel project at the same company: an open-source privacy filter that strips sensitive patient data from documents before any LLM sees them. Alongside: 20+ Power BI dashboards across eight departments, four peer-reviewed clinical papers (corresponding author on one), and a couple of trading bots running unattended on a VPS. My route here ran through a math-heavy MSc and several years translating between doctors and data, which is probably why working with language models felt less like a new skill and more like the place this kind of thinking was headed.

Download CV Get in touch
nick.ashouri9248@gmail.com
Arnsberg, Germany
gh github.com/nick9248
in linkedin.com/in/nick-ashouri

Experience

Tehran Municipality

IT Technician
  • Led the Smart City initiative for Tehran's District 6 — instrumented public electricity and water systems and built Elasticsearch and Kibana dashboards on top of SQL pipelines so district staff could see energy usage patterns directly, driving a 10% reduction in resource consumption and shrinking reporting cycles by 20%.
  • Ran tier 1–3 technical support for the district's IT estate over 3.5 years alongside CS studies, raising system uptime by 15%.
SQLElasticsearchKibanaSmart City

reha assist GmbH

Application Developer
  • Established the foundational BI infrastructure for a rehabilitation management organisation of 100+ employees — configured Enaio DMS document workflows and built the first-generation Power BI reports that all subsequent dashboards built upon.
  • Designed and implemented ETL pipelines connecting Navision ERP data to the reporting layer, setting the data architecture standards used across all later projects.
Enaio DMSPower BIETLNavision ERP

reha assist GmbH

LLM Engineer & BI Developer
  • Built and deployed an Azure Functions pipeline using OpenAI models and LangChain to classify, index, and summarise clinical case files against Enaio DMS — processing ~100 medical documents daily at ~90% average accuracy, collapsing per-document handling from ~30 minutes to seconds.
  • Tiered model selection by task: a lightweight model for OCR preprocessing and a reasoning model for clinical content — balancing throughput, cost, and quality at production scale rather than defaulting to one model end-to-end.
  • Delivered 20+ Power BI dashboards across sales, HR, operations, and executive reporting — consolidating Navision ERP, Enaio DMS, and HR sources into live views, including a bonus calculation reduced from 3 days to instant.
  • Working alongside the IRU research institute (Dr. Schmidt) — translated clinical questions from doctors and department heads into SQL pipelines, dashboards, and AI workflows, contributing to 4 peer-reviewed publications, including corresponding authorship on a 408-patient outcomes study (FL_Score −56%, Cohen's d=1.37, p<0.001).
  • Currently developing a patient-data privacy filter built on an open-source LLM to redact sensitive identifiers locally before any cloud inference — privacy-by-design for clinical AI.
OpenAI APILangChainAzure FunctionsPower BIDAXSQL Server

Selected Projects

Options Trading DSS

2026 – Present

Automated decision-support platform for BTC/ETH options on Deribit — ML-based regime detection (LightGBM), displacement scanner with conviction scoring, on-chain analytics (GEX/DEX, max pain, IV surface). Deployed 24/7 on Hetzner VPS with Telegram alerts.

Polymarket Copy-Trading Bot

2026 – Present

Blockchain trade-mirroring bot using EIP-712 signed CLOB orders on Polymarket's prediction market. Multi-layer safety guardrails: geo-verification, slippage guard, genesis block protection, and on-chain transaction validator.

On-Chain Options Dashboard

2025 – Present

Local web dashboard for BTC/ETH on-chain options analytics — GEX/DEX exposure, max pain, implied volatility surface, and block trade feed. Flask REST API backend with a React frontend.

Publications

Corresponding Author

Effectiveness of Telephone-Based Treatment Accompaniment in Musculoskeletal Rehabilitation: A Paired Longitudinal Analysis Using ICUC Scores

1st Co-Author

Detecting disorders early in the healing process: Telephone support during the healing process using the ICUC score

Co-Author

Long-term care of polytraumatized patients, requirements for rehabilitation management — Part 2: After the 78th week

Numbers & Impacts

0
Medical docs processed daily
Each working day, ~100 medical documents arrive — diagnoses, procedure reports, invoices, referrals. Previously each took ~30 minutes to classify and route manually. The Azure Function App now handles the full pipeline in seconds: OCR, classification, indexing, and case-summary extraction.
0
classification accuracy
Average accuracy across all document types — Verordnungen, AU certificates, surgical reports, invoices. Achieved through deliberate model selection: GPT-4.1-nano for OCR pre-processing, GPT-4.1 for complex reasoning and clinical extraction.
0
Power BI dashboards delivered
20+ dashboards delivered across 8 departments — executive, sales, HR, operations, IT. Integrating Navision ERP, Enaio DMS, and HR data into one real-time view. Includes a bonus calculation reduced from 3 days to instant.
0
Peer-reviewed publications
Four peer-reviewed publications in healthcare and rehabilitation research — one published in Die Unfallchirurgie (Oct 2025), three under review. Corresponding author on a 408-patient outcomes study (FL_Score −56%, Cohen's d=1.37, p<0.001). Full pipeline: SQL extraction, statistical analysis, clinical reporting.

hover a number to see the story behind it

Languages

US
US Flag
English
Professional Working Proficiency
DE
German Flag
German
B1 · TELC Certified
Advancing to B2
IR
Iranian Flag
Persian
Native

Skills

LLM Engineering
OpenAI APILangChainAzure OpenAIAzure FunctionsPrompt EngineeringDocument ProcessingRAGNLPHuggingFaceOllamaPII Redaction
Business Intelligence
Power BIDAXPower QueryDataflowsNavision ERPEnaio DMSETLData ModelingSQL Server
Programming
PythonMachine LearningClassificationRegressionLightGBMXGBoostscikit-learnpandasNumPyPostgreSQLGitREST APIsWeb3EIP-712
Research & Analysis
Statistical AnalysisLongitudinal StudiesCohen's dWilcoxonBootstrap CI
Leadership
Requirements TranslationStakeholder ManagementTeam CoachingTechnical DocumentationProcess Optimization

Education

BSc Computer Science

Kharazmi University, Tehran · 2013 – 2018
  • Foundations in algorithms, data structures, software engineering, and discrete mathematics.
  • Thesis and coursework spanning operating systems, database design, computer networks, and software architecture.

MSc Computer Science

Bergische Universität Wuppertal, Germany · 2019 – ongoing
  • 3 courses remaining — pursued alongside full-time employment in AI engineering and BI development.
  • Focus on advanced algorithms, machine learning foundations, and distributed systems.
  • Academic context that directly informs the statistical methods applied in three peer-reviewed publications.

Certifications

Hover any node to see the certificate name.

Beyond Work

Football
Real Madrid

It has been over 20 years that I am playing football. Pro and amature. In the recent years, knee injury (and laziness) held my from playing consistently and on top level. currently i am playing for Tus Wickede II team.

Billiards
Ronnie O'Sullivan

More details coming soon.

Cinema
Interstellar

More details coming soon.

Swimming
Michael Phelps

More details coming soon.

Gym
Arnold Schwarzenegger

More details coming soon.

Gaming
Kratos

More details coming soon.

Nick Ashouri  ·  AI & LLM Engineer