I’m a physicist/PhD-turned Data Scientist who likes problems where data is messy, decisions matter, and “good enough” needs to be measurable. My background at CERN shaped how I work: start with a baseline, validate hard, quantify uncertainty, and build workflows that others can trust.

Right now I’m looking for Data Scientist / Applied ML roles (production DS or ML) where I can own problems end-to-end: data → model/analysis → evaluation → delivery.

• Turning noisy real-world data into decision-ready metrics and models
• Validation-first modeling (leakage checks, time splits, backtests, error analysis)
• Building robust pipelines (automation, reproducibility, clear handoffs)
• Handling uncertainty (intervals, confidence bounds, risk-aware decisioning)
• Communicating results so stakeholders can act (trade-offs, thresholds, guardrails)

• Adset profitability scoring (pLTV vs CAC) — weekly budget steering under incomplete attribution
  Dynamic 3–8 week lookback + binomial fallback; ~500–1000 adsets/week; contributed to ~20% ROI improvement
  👉 https://github.com/nakolkar16/adset-profitability-scoring

• LinkedIn: https://www.linkedin.com/in/nakolkar/
• Email: nilimaakolkar16@gmail.com