Key Responsibilities

- Perform data preparation, feature engineering, and exploratory analysis on transactional, behavioural, third-party, and alternative data.

- Develop predictive models for risk scoring, underwriting, propensity, early-warning signals, forecasting, and workflow prioritisation.

- Define target variables, observation windows, and performance windows carefully, accounting for time lags, maturity, and data leakage.

- Validate models using out-of-time testing and assess performance on discrimination, calibration, stability, and business impact.

- Help turn model outputs into practical decision rules, thresholds, segments, and action strategies.

- Support model deployment, monitoring for drift and performance, and recommend recalibration or retraining when needed.

- Partner with engineering to keep training and production features consistent and to support reliable pipelines.

- Maintain transparent documentation, reproducibility, and explainability so models can support governance and review.

- Contribute to AI/LLM initiatives (e.g., virtual CFO or agent-style workflows) through feature design, evaluation, or integration as needed.

Required Skills and Experience

- Strong proficiency in Python and SQL, with experience working on large datasets.





- Practical experience with logistic regression and tree-based models (XGBoost, LightGBM, CatBoost, or similar), and knowing when simpler models are preferable.

- Solid grounding in statistics and experimentation, including sampling and selection bias, class imbalance, and careful treatment of leakage.

- Experience evaluating models using AUC, KS, Gini, precision/recall, calibration, and stability measures such as PSI.

- Familiarity with explainability techniques such as SHAP and ability to explain model behaviour to non-technical stakeholders.

- Experience contributing to model deployment and monitoring in production environments.

- Bachelors or masters degree in statistics, mathematics, computer science, engineering, economics, or a related quantitative field.

Nice to Have

- Exposure to deploying or evaluating AI/LLM inference workflows (latency, cost, reliability, output quality).

- Familiarity with LLM tool/function calling, external API integration, or multi-step agent-style workflows.

- Understanding of supervised fine-tuning, instruction tuning, or preference optimisation approaches.

Experience putting evaluation, guardrails, observability, or human review in place for sensitive AI

📌 Data Scientist- LLM/GenAI (Noida)
🏢 Huey Tech
📍 Noida