Chitra Varma

I'm a Computer Science graduate from Pillai HOC College with a strong academic record (93.1%, 2nd position in first year) and a portfolio built around one core skill: turning ambiguous, real-world data into decisions that hold up under scrutiny.

My work spans the full data science lifecycle. On the experimentation side, I built an Experiment Decision Engine — a platform that takes raw A/B test data and produces ship/no-ship recommendations using frequentist two-proportion tests, Bayesian Beta-Binomial posteriors, and expected-loss decision rules. It includes pre-launch power analysis, Sample Ratio Mismatch guardrails, sequential testing with alpha-spending to control false positives from continuous monitoring, and CUPED variance reduction — which cut outcome variance by ~10% and moved a borderline result from p=0.045 to statistical significance. It's covered by 21 property-based tests, including an A/A false-positive check.

On the causal inference side, I built a hospital readmission project that goes beyond risk prediction: a calibrated LightGBM model estimates who's likely to be readmitted, but a T-learner uplift model estimates who actually benefits from intervention — a meaningfully different (and more actionable) question. I then formulated the limited follow-up capacity as an integer linear programming problem to maximize prevented readmissions under a fixed budget, showing that benefit-based targeting outperforms risk-based targeting at equal cost.

I've also worked in NLP and computer vision — building InterViewWise, an interview analytics platform that scores candidate responses using speech-to-text, GPT-4o-mini evaluation, sentiment analysis, and facial/pose landmark tracking, validated through a structured 12-user survey. On the applied ML side, I've deployed a regression-based real estate price predictor (Flask, 13,000+ listings) and built a Power BI dashboard that identified $22.5K in discount-driven margin leakage, backed by DAX time-intelligence measures.

Before this, I interned as a Java Developer at Ganiska Technology, where I designed a MySQL database with Python/REST integration, built ETL pipelines, and implemented role-based access control — solid grounding in the engineering side of data work, not just modeling.

I'm particularly drawn to roles involving experimentation, causal inference, and decision science — problems where statistical rigor directly shapes what a company does next. I'm comfortable with ambiguity, enjoy translating technical findings for non-technical stakeholders, and I'm currently expanding into Azure ML, TensorFlow/CNNs, and PySpark to round out my deployment and scale skills. Open to remote roles globally.