Business Analytics · AI & Data
MS Business Analytics graduate (with Distinction) from Central European University. The common denominator across my career is staying grounded in how businesses actually operate: inside their systems, behind their numbers, and within their regulatory environments. Backed by nearly two decades in financial ops, tech, and law, I am seeking analytics roles where rigorous thinking meets modern data tooling.
I recently completed an MS in Business Analytics at Central European University, graduating with Distinction (3.81 / 4.0 GPA) and receiving the Outstanding Capstone Project Award for my cohort. My academic work spans deep learning, time series forecasting, causal inference, geospatial analysis and data engineering, culminating in a capstone for an industry sponsor where I built and benchmarked prompt-injection defenses for enterprise AI agents.
Before CEU, the common thread throughout my career has been staying grounded in how businesses actually operate: inside their systems, behind their numbers, and within their regulatory environments. Across nearly two decades in financial operations, tech, and law, I have owned reporting and forecasting on $50M–$200M portfolios, developed custom SQL reporting solutions for enterprise accounting systems, and led document-review teams as a California-admitted attorney.
I am drawn to analytics and data roles in finance and regulated industries where quantitative rigor, business instinct, and modern tooling come together to solve complex operational problems.
Central European University
MS in Business Analytics
Graduated with Distinction · GPA 3.81 / 4.0 · Outstanding Capstone Project Award
Coursework: deep learning, time series forecasting, causal inference, geospatial analysis, data engineering, data analytics with AI, data governance / Responsible AI, managing data science teams, designing analytics projects, financial management, fintech.
UC Hastings College of the Law
JD
Admitted to the California Bar, June 2009 (#264084)
Best Oral Advocate and Best Appellate Brief (Honorable Mention), Moot Court; Legal Writing & Research teaching assistant.
Cornell University
BA, Classics
Graduated magna cum laude
Frederick Conger Wood Fellowship · Dean's List · Dickerson Prize for best first-year essay (Dartmouth, before transfer) · President of Telluride House (administered a $150K annual scholarship-house budget).
A deployed retrieval-augmented question-answering assistant over the FATF Recommendations, the global rulebook for how banks must fight money laundering and terrorist financing (set by the Financial Action Task Force and adopted by more than 200 countries). It answers only from the source text with inline page citations, abstains when a question is out of scope, treats retrieved text as data rather than instructions, and reports measured retrieval and answer-faithfulness metrics against a hand-built gold set. Reuses the prompt-injection and LLM-judge evaluation methodology from my MS capstone.
Built and benchmarked input- and output-side prompt-injection defenses for enterprise AI agents: transformer classifiers (DeBERTa-v3, Prompt Guard 2), an agent-plus-LLM-judge pipeline, and LoRA fine-tuning. Delivered a Business Decision Framework mapping deployment scenarios to defense stacks and human-in-the-loop thresholds that minimize expected cost.
An interactive Tableau dashboard testing how well The Economist's Big Mac Index tracks official price statistics, built on a Python (pandas) pipeline drawing from The Economist, Eurostat price levels, and Hungarian consumer prices (KSH). The index ranks countries' price levels well but understates the spread between them, overstates short-run inflation, and hides wide variation inside the euro area — a look at where a popular proxy holds up and where it breaks.
Forecast monthly CVE disclosures from 314,774 NIST NVD records (2005–2025). SARIMA reached 15.17% MAPE on a 36-month holdout, beating an AutoGluon ensemble (21.41%) and the seasonal naive baseline (21.36%). Identified a 2017 reporting-coverage structural break.
Built a composite wildfire vulnerability index for LA County on an H3 hex grid (~460 m), combining fire exposure, evacuation capacity, and recovery risk from nine geospatial and census sources. Used local Moran's I to surface three high-vulnerability profiles; back-tested against the Woolsey, Eaton, and Palisades fires (39–72% top-quintile coverage vs a 20% baseline).
Estimated the effect of R&D intensity on total factor productivity across a 94-country panel (2000–2019) using two-way fixed effects, first-difference, and instrumental-variable specifications with clustered standard errors. Found a robust near-zero association and transparently documented two instruments that failed first-stage relevance.
As primary developer, built the multi-agent orchestration and ChromaDB RAG pipeline: an orchestrator agent profiles the user and routes to a retrieval agent over a vector store of TMDB metadata. Deployed a streaming Chainlit app with multi-turn memory, serving models via AWS Bedrock through LiteLLM.
As primary developer, built an end-to-end pipeline scraping English, German, and Hungarian news, translating non-English text with AWS Translate, and scoring sentiment with AWS Comprehend using a custom chunk-and-aggregate algorithm. Stored processed records as structured JSON in S3 for longitudinal study.
Built a bronze, silver, gold pipeline integrating Chicago business-license (CSV) and NYC building-permit (JSON) data into a clean, validated analytical layer. Caught 247 records with impossible date relationships and imputed them with audit-flag columns for traceability, and used the DuckDB spatial extension for distance-from-city-center analysis, plus business KPIs.
Designed a three-part toolkit to push Vienna's material footprint toward 7 tons per capita by 2050: a transparency app surfacing food footprints to individuals, procurement reform for the ~100,000 public meals served daily, and a tiered support framework for community food cooperatives.
Programming & Querying
PythonSQLGitDocker
Machine Learning & AI
scikit-learnPyTorchTensorFlow / KerasHuggingFaceLoRA / PEFTRAG & multi-agent
Statistics & Econometrics
RegressionCausal inference (TWFE / DiD / IV)A/B testingTime series (SARIMA)
Cloud & Data Engineering
AWS (Bedrock / Translate / Comprehend / S3)dbtSnowflakeDatabricksDuckDBChromaDB
Visualization & Reporting
Power BITableauExcel (VBA, modeling)Salesforce
Regulatory
JDCalifornia BarEU directive analysis
Languages
English (Native)Hungarian (Native)German (A2, in progress)
Build Group Construction Company Inc.
Division Controller
El Toro Holdings
Project Manager
Urban Developments, LLC
Operations Manager
Contract Attorney — O'Melveny, Manatt, Ropes & Gray, Cappello & Noël
Litigation & Document Review
Covington & Burling LLP
Legal Intern
Yardi Systems, Inc.
Computer Programmer / Software Developer
Open to analyst and analytics roles across the EU. EU citizen (Hungary), with full right to work throughout the EU.