M.Sc. student in Applied Data Science & AI at SRH University, based in Regensburg, Germany. I build machine-learning systems end to end: from messy public data to a model that is honestly evaluated and a tool someone can use.
🔎 Looking for a working-student role (Werkstudent) in data science, machine learning or applied GenAI, in Regensburg or remote: up to 20 h/week during the semester, full-time in semester breaks.
| Project | What it does | Highlights |
|---|---|---|
| Document Q&A Assistant (RAG) Oct 2026 |
Answers questions about your own PDFs and notes from the passages it retrieves, cites them, and refuses to guess. Claude, OpenAI, Ollama or a free local LLM. | Evaluated on 1,000 SQuAD 2.0 questions: right passage ranked 1st for 88% (hybrid BM25 + FAISS + cross-encoder re-ranking) · refuses 93% of off-topic questions · 3 bugs found by its own evaluation · 22 unit tests |
| NLP News Intelligence Jun 2026 |
Classical NLP vs. Sentence-BERT vs. LDA on 127,600 AG News articles: classification, semantic search, topic discovery. | 91.2% accuracy · semantic search P@5 0.844 vs. 0.776 keyword |
| Motor Insurance Claims Copilot M.Sc. group project · Jun 2026 |
Audit-safe GenAI copilot for motor insurance claims: schema-constrained extraction, policy-scoped hybrid retrieval (BM25 + dense) and deterministic rule engines. | Human-in-the-loop (the AI never decides) · abstains on conflicting evidence · source-span citation checks · hash-chained audit log · 8-stage evaluation incl. prompt injection & fairness |
| Project | What it does | Highlights |
|---|---|---|
| German Electricity Price Forecasting Oct 2026 |
Forecasts tomorrow's hourly day-ahead price for Germany/Luxembourg from 8 years of public SMARD data, using only information available before the auction. | 26% lower error than the naive forecast (MAE 22.0 vs 29.7 EUR/MWh) · 8,712-hour rolling-origin backtest · leakage tests |
| COVID-19 ICU & Mortality Forecasting M.Sc. project · 2026 |
Forecasts ICU occupancy and deaths for 4 regions, comparing Linear, Random Forest and XGBoost with leakage-free time-series validation. | −87% MAE vs. linear baseline (Germany) · found the Omicron regime shift in the residuals |
| Gut Transit in ICU Trauma Patients (Biostatistics) M.Sc. project · Mar 2026 |
Statistical replication of Rauch et al. (2012): SmartPill capsule transit times in 8 ventilated trauma patients vs 87 healthy volunteers. | Gastric emptying 4.5× slower (13.9 h vs 3.0 h, Mann–Whitney p < .001) · χ²/Fisher, ANOVA, distribution fitting · showed Pearson vs Spearman changes the conclusion |
| EMG Muscle Fatigue Classification M.Sc. project · Jun 2026 |
sEMG signal processing + ML for fatigue detection in unseen people: 9,540 windows from 14 participants, 10 time- and frequency-domain features. | Leave-one-subject-out · mean F1 0.416 vs 0.625 for a trivial baseline: features differ by group, but don't yet transfer across people |
| Project | What it does | Highlights |
|---|---|---|
| EU Safety Gate Motor Vehicle Recalls Dashboard Oct 2026 |
Power BI dashboard on 4,184 motor-vehicle recall alerts from the EU Safety Gate system (2016–2026): brands, components, hazards and the countries affected. | Every card matches an independent Python recount · 9 data issues found and fixed · Power Query, DAX |
More write-ups on my portfolio.
- Languages: Python, SQL, Java, C, Bash
- Machine learning: scikit-learn, XGBoost, time-series forecasting, model evaluation & cross-validation
- NLP & GenAI: RAG, hybrid retrieval (BM25 + dense), sentence-transformers, FAISS, OpenAI API, Hugging Face Transformers, spaCy, NLTK, gensim (LDA)
- Data & signals: pandas, NumPy, SciPy, ETL pipelines, signal processing (sEMG)
- Statistics: statsmodels, JASP, hypothesis testing
- Visualisation & BI: matplotlib, seaborn, Power BI (Power Query, DAX), Tableau, Excel
- Databases & tools: PostgreSQL, MySQL, SQLite, MongoDB, Git, Docker, Linux, Jupyter, Streamlit
- Data Science Intern, TechnoTut, Mysuru, India (Feb – May 2025) Built and evaluated supervised and unsupervised ML models in Python; presented results to technical and non-technical stakeholders.
- Data Analysis Trainee, SPORTS-KPI, India (Sep – Dec 2024) Built KPI dashboards and reports for football and kabaddi match data.
- M.Sc. Applied Data Science & AI, SRH University (2025 – 2027, expected)
- B.E. Computer Science & Engineering, The National Institute of Engineering, Mysuru (2023 – 2025), First Class
- Diploma in Computer Science, JSS Polytechnic for Women, Mysuru (2019 – 2022), First Class with Distinction
Languages: English (C1) · German (B1) · Kannada (native)