Junior AI Engineer

Responsibilities

You will own scoped model workstreams including data preparation, training, evaluation, and error analysis under senior guidance. Additionally, you will collaborate with radiologists to define labels and support the deployment of validated models into production.

Requirements

Candidates must hold a degree from a selective institution in a quantitative field and possess 2-4 years of hands-on deep learning experience. Strong proficiency in Python, PyTorch or TensorFlow, and a proven track record of building computer vision systems are required.

Skills

• Python
• PyTorch
• TensorFlow
• Keras
• NumPy
• Deep learning
• Computer vision
• CNNs
• Vision transformers
• Detection
• Segmentation
• DICOM
• Clinical metrics
• Docker
• FastAPI
• Git

Benefits

• Mentorship
• On-premise GPU infrastructure
• Clinical collaboration

About this role

About Teleradiology Solutions / Daignostix
Teleradiology Solutions (TRS), founded in 2002 by two Yale-trained physicians and headquartered in Ardmore, PA, provides 24x7 diagnostic radiology reporting to over 150 hospitals across 21 countries, through a network of radiologists (US, Indian certified) and technology-driven workflows.
Daignostix, the AI arm of TRS based in Whitefield, Bengaluru, builds clinical AI for medical imaging and healthcare workflows. TRS also maintains an active academic and research focus, including AI in radiology (www.radguru.net).

Role Summary:
Daignostix builds clinical AI for CT, MR, X-ray and ultrasound, with algorithms spanning neuro and stroke, pulmonary embolism, aortic pathology, urology, spine and rib fracture, solid-organ trauma and paediatric chest. Our models are used alongside radiologists (US, Indian certified), so quality is judged by clinical performance and reliability, not benchmark scores.
As a Junior AI Engineer, you will own scoped model workstreams (data, training, evaluation and error analysis) under the guidance of senior AI engineers, and support the move of validated models into production. Medical-imaging experience is welcome but not required; serious hands-on deep-learning work is.

Roles and Responsibilities:
● Model development. Train, fine-tune and evaluate CNN- and transformer-based models for classification, detection and segmentation, starting from strong baselines and running disciplined, reproducible experiments.
● Scoped ownership. Own a defined model workstream under senior review, from dataset preparation through training, validation, documentation and hand-off.
● Data pipelines. Build and maintain DICOM pipelines for ingestion, de-identification, series selection, pre-processing, augmentation, annotation handling and dataset versioning.
● Evaluation and error analysis. Report clinical metrics (sensitivity, specificity, PPV/NPV, ROC-AUC/PR-AUC, per-case and per-lesion), analyse false positives and negatives by subgroup, and explain why the model fails.
● Radiologist collaboration. Work with radiologists (US, Indian certified) to define labels, review difficult cases and turn clinical feedback into measurable model improvements.
● Production support. Contribute to packaging validated models as tested inference services (Python, Docker, FastAPI), including unit and regression tests that confirm changes do not alter clinical outputs. Senior engineers lead serving architecture, optimisation and deployment.

Requirements:
Must-have:
● Education. B.Tech/B.E./M.Tech/MS/M.Sc. from an IIT, IISc, IIIT, BITS Pilani or a comparably selective institution, in Computer Science, Electrical/Electronics Engineering, AI, Data Science, Mathematics, Statistics or a related quantitative field, with a strong academic record.
● Experience. 2-4 years of hands-on AI/ML engineering or research experience, including deep-learning work beyond coursework, tutorials, notebooks or API wrappers.
● Proof of building. At least one substantial computer-vision system or research contribution that you personally drove into a reproducible pipeline, service or real downstream use, and can defend in technical depth.
● Programming. Strong Python and NumPy, and deep hands-on experience in PyTorch or TensorFlow/Keras. Able to debug data loaders, leakage, incorrect metrics and unstable training.
● Fundamentals. Solid grounding in deep learning and computer vision (CNNs, vision transformers, detection, segmentation, transfer learning) and in statistics and experimental design (ROC/PR analysis, thresholds, calibration).
● Engineering basics. Linux, Git, clean modular Python, Docker, pytest and basic CI/CD.
Good-to-have:
● Medical imaging. Experience with DICOM/NIfTI, MONAI or nnU-Net.
● Production exposure. Model serving with FastAPI, ONNX/TensorRT, and experiment tracking with MLflow or Weights & Biases.
● LLMs and cloud. Exposure to LLMs, embeddings or RAG, and to one of AWS, Azure or GCP. These add to, but do not replace, core deep-learning skills.

Key Skills:
Python | PyTorch | TensorFlow / Keras | NumPy | OpenCV | scikit-learn | CNNs | Vision Transformers | Detection | Segmentation | DICOM / NIfTI | Clinical Metrics | Error Analysis | Git | Linux | Docker | FastAPI | pytest | MLflow / Weights & Biases

Why Join Daignostix:
● Real clinical AI. Build models for real medical-imaging workflows and see how they affect radiologists and patients.
● Direct access to radiologists. Work closely with subspecialist radiologists (US, Indian certified) on ground truth, difficult cases and failure analysis.
● Mentorship with ownership. Senior technical review, dedicated on-premise GPU infrastructure, and visibility across development, deployment, clinical validation and regulatory pathways.

About Telradsol

Hospitals and Health Care · 501-1,000 employees · headquartered in Bangalore · founded 2002