Machine Learning Engineer

Hula Global is hiring a Machine Learning Engineer in Bangalore to build and operate models using Oolka's data. The work includes repayment-likelihood scoring, negotiation outcome prediction and credit-risk signals. The posting lists 3–6 years of experience and specifies 3+ years building and shipping ML models in production, rather than only integrating third-party AI APIs.

Responsibilities

• Own the model lifecycle from data collection and feature engineering through training, validation, deployment and monitoring. Build feature pipelines using credit bureau, transaction and repayment data.
• Fine-tune LLMs and smaller models for domain-specific tasks, including structured extraction from credit reports and negotiation dialogue quality.
• Build evaluation frameworks to catch quality regressions before release. Operate model serving with batching, quantisation, versioning and rollback; monitor production drift, degradation and bias, and own retraining.
• Partner with the AI engineering team on improving models and serving them in the live product.

Requirements

• Hands-on model training and fine-tuning with PyTorch or TensorFlow, including classical ML and/or LLM fine-tuning; strong feature engineering and structured or tabular data pipeline experience.
• Experience with model-serving frameworks such as Triton, TorchServe or TensorFlow Serving; inference optimisation through batching, quantisation and distillation; and MLOps practices including experiment tracking, model registries and model CI/CD, using tools such as MLflow, Kubeflow or SageMaker.
• Experience shipping models that predict real-world outcomes, such as risk, churn or ranking; offline and online evaluation using held-out test sets, A/B testing and shadow deployment; and diagnosing production model-quality regressions to their data or training cause.
• Understanding of LoRA/PEFT, when fine-tuning is preferable to prompting, and the bias, fairness and explainability considerations of credit decisions. Credit, lending or fraud experience is a strong plus. The posting's structured metadata lists a bachelor's degree.