I presently work part-time as a research assistant at the Chair of Technology and Innovation Management at ETH Zurich, under the supervision of Prof. Dr. Stefano Brusoni. My work focuses on the intersection of machine learning, data engineering, and technical analysis, with a strong emphasis on automating complex workflows and enhancing data-driven decision-making processes.

Enterprise ML & Technical Analysis Projects

March 2024 - January 2025

Led the data engineering and analytical components of a large-scale research collaboration between ETH Zurich and Zühlke, investigating machine learning and AI adoption patterns across 600+ enterprises in Europe and North America. This comprehensive study spanned multiple industries including consumer discretionary, IT, manufacturing, financial services, and healthcare sectors.

View Exec Summary
Download Full Report (PDF)

Technical Implementation

  • End-to-End Data Pipeline: Architected a comprehensive data processing framework integrating multiple sources (Prolific survey responses, Qualtrics data, demographics) with robust cleaning, standardization, and anonymization processes
  • Statistical Analysis Engine: Developed an automated testing framework executing 10,000+ statistical tests (t-tests, ANOVA, Kruskal-Wallis, Chi-Square, Levene’s, Mann-Whitney U) with intelligent test selection based on data characteristics
  • NLP & AI Integration: Implemented text classification pipelines utilizing OpenAI’s LLM APIs for categorization of free-text responses and sentiment analysis
  • Dynamic Visualization System: Created a visualization engine generating 2,600+ charts automatically selected based on variable types and distributions
  • Automated Reporting: Engineered systems to transform complex statistical outputs into business-friendly formats through programmatic Excel reporting

Technologies

# Core Data Stack
import pandas as pd
import numpy as np
import scipy.stats as stats
import statsmodels.api as sm

# Visualization
import matplotlib.pyplot as plt
import seaborn as sns

# NLP & ML
from openai import OpenAI

Impact & Outcomes

  • Reduced manual analysis time by 80% through comprehensive automation workflows
  • Identified statistically significant ML adoption patterns across regions, company sizes, and industry sectors
  • Created a reproducible research framework now used for ongoing studies at the Chair
  • Generated actionable insights on ethical considerations, operational efficiency, and strategic ML/AI integration that informed the final research publication

Unicode Technical Consortium Document Analysis

Developed a comprehensive system to analyze 20,000+ technical documents from the Unicode Technical Consortium spanning a decade of standardization history. This project enabled identification of key technological trends, particularly in emoji standardization and adoption patterns.

View on GitHub

Technical Implementation

  • Scalable Web Crawling: Engineered concurrent HTTP request handling with automatic retry logic and connection pooling, achieving 94% successful download rate for heterogeneous document sources
  • Multi-format Document Processing: Built extraction engines for PDF, HTML, and plaintext documents with format-adaptive parsers handling the evolution of UTC’s document repository structure
  • Document Classification System: Developed a hierarchical classifier with 30+ document categories using domain-specific keyword patterns and regex-based boundary matching
  • Keyword Extraction Framework: Implemented parallel processing using multiple algorithms (RAKE-NLTK, YAKE) for comprehensive content summarization
  • LLM Integration: Engineered configurable OpenAI API client with token optimization techniques reducing API costs by 42% while maintaining extraction quality

Technologies

# Web Scraping & Document Processing
import asyncio
import aiohttp
from bs4 import BeautifulSoup
import PyPDF2

# NLP & Text Analysis
import nltk
from rake_nltk import Rake
import yake
from sklearn.feature_extraction.text import TfidfVectorizer

# LLM Integration
from openai import OpenAI

Impact & Outcomes

  • Created a comprehensive emoji reference database mapping 5,000+ emoji representations across different notation systems
  • Enabled identification of cyclical patterns in Unicode standardization process
  • Developed attribution analysis identifying key contributors and their influence over time
  • Built visualization packages for temporal trends, category distributions, and emoji adoption patterns
  • Reduced overall project costs by 31% through intelligent API usage and batch optimization

Patent Analysis for Technological Shift Detection

Designed and implemented an advanced system for analyzing patent data from semiconductor companies to identify technological shifts and innovation patterns. This project processed 10,000+ patents across multiple decades to reveal strategic technology adoption trends in the semiconductor industry.

View on GitHub

Technical Implementation

  • High-Performance Data Acquisition: Built an asyncio-powered concurrent scraping system with intelligent session management, achieving 20x faster processing compared to sequential approaches
  • Text Optimization Pipeline: Engineered token optimization for 15,000+ patent documents to fit within embedding windows while preserving critical technical information
  • Topic Modeling: Deployed BERTopic with Sentence Transformers embeddings to identify 80+ distinct technology clusters across 20+ years of semiconductor industry patents
  • Domain-Specific Filtering: Implemented LLM integration using OpenAI models to distinguish semiconductor-specific patents from general technology with 95% accuracyResearch Assistant at Chair of Technology and Innovation Management under Prof. Name*
    August 2023 - December 2023
  • Technological Shift Detection: Created visualization and analysis of topic trends over time, identifying 5 major technological shifts in the semiconductor industry

Technologies

# Concurrent Processing
import asyncio
import aiohttp

# NLP & Embeddings
from sentence_transformers import SentenceTransformer
from bertopic import BERTopic
import umap
import hdbscan

# GPU Acceleration
import torch

# LLM Integration
from openai import OpenAI

Impact & Outcomes

  • Enabled identification of emerging technological areas and company-specific innovation patterns
  • Created time-series analysis of patent trends through temporal tracking of topic distributions
  • Developed a modular, configuration-driven pipeline that scaled from single-company to industry-wide analysis
  • Produced visual representations of technological evolution in the semiconductor industry
  • Built a reusable framework for patent analysis that can be applied to other technology sectors

License

This project is licensed under the MIT License.

Updated:

You May Also Enjoy