About the Role :

Solving for the next billion users globally, True Credits is tackling complex problems in a rapidly growing fintech domain using massive data pipelines and creative solutions across Growth, Lead Generation, Underwriting, Collections, and beyond.

We are searching for a critical thinker with experience in machine learning workflows, passionate about solving business problems using data and thriving in a dynamic, creative, and collaborative environment.

This is not a role where your work finishes with a Jupyter Notebook, tweaking hyperparameters, and throwing a pickle file over the wall to DevOps. At True Credits you build resilient AI systems that won't melt under millions of concurrent production requests.

The Foundations :

- Strong background in fundamentals : concurrency, distributed systems, memory management, data structures, and system design.

Tech Stack :

- 5+ years of experience shipping production-grade code.

- Deep mastery of Python and either Go, Java, or C++.

- ML Infrastructure : Hands-on experience with production ML serving frameworks (Triton, TorchScript, BentoML) and MLOps platforms (MLflow, Kubeflow).

- You know how to build CI/CD for ML.

- You don't just write code; you write automated tests for model behavior, input validation, and fallback mechanisms.

Roles and responsibilities :

- Architect Inference Infrastructure : Own the end-to-end performance of our LIP execution layer.

- You will design and build highly scalable microservices that serve models, ensure predictive accuracy, and maintain a strict execution budget per request.

- Engineer Real-Time Feature Streaming : Design, optimize, and maintain distributed, event-driven data pipelines (using Flink, Kafka, Spark) to extract and aggregate alternative data features on the fly.

- You will turn chaotic user activity streams into clean, structured inputs for our models in real time.

- Build Next-Gen MLOps Ecosystems : Develop automated, robust CI/CD frameworks specifically for machine learning.

- This includes building self-healing pipeline abstractions, automated model retraining loops, and automated shadow-testing frameworks where new models can safely run alongside production models before a full rollout.

- Eliminate Drift and Data Skew : Implement rigorous monitoring, alerting, and automated validation systems to catch training-serving data skew, input anomalies, and feature drift before they impact underwriting or risk decisions.