Data Scientist- 5Plus Years ( Fine Tuning)

AI/ML Engineer (5+ Years)

Were building AI systems that convert unstructured visual data into structured, usable outputs that power real-world workflows.

This role focuses on solving messy, high-variability problems and delivering reliable systems in production, not just training models in isolation.

What Youll Do

Build and improve end-to-end AI/ML pipelines (input preprocessing inference structured output)Work with noisy, inconsistent datasets and drive improvements in data quality and model performanceDevelop preprocessing pipelines (e.g., normalization, transformations, input standardization)Run experiments, tune models, and evaluate trade-offs across accuracy, latency, and costCollaborate with engineering teams to productionize models (APIs, batch/real-time systems)What Were Looking For

5+ years of hands-on experience in ML/AI with exposure to computer visionStrong Python skills with production-quality coding practicesSolid understanding of microservices and system design.Experience building and deploying ML systems beyond experimentationExperience working on systems that extract or process information from complex visual inputs (e.g., documents, forms, or similar structured outputs)Strong understanding of neural network architectures (e.g., CNNs, Transformers, sequence models) and their real-world trade-offs.Ability to:Select appropriate approaches based on problem constraintsFine-tune and optimize models for performance and reliabilityDebug model behavior and failure cases in productionStrong understanding of evaluation metrics, error analysis, and model limitationsGood to Have

Experience with cloud environments (e.g., Azure or similar)Familiarity with scalable data pipelines, containerization, or distributed systemsExposure to multi-modal systems (vision + text)What We Dont Want

Only academic or demo project experienceOver-reliance on plug-and-play APIs without understanding internalsInability to explain past work end-to-end (data model output impact)Profiles focused only on generic image classification or object detection use cases without real-world deployment experience .