Artificial Intelligence Researcher

AI for Early Cancer Detection (Computer Vision)

Location: On-site — Kochi, Kerala, India  

Employment: Full-time (6–12 month contract with potential extension)  

Start date: Immediate

## About the project

We are building a pilot AI system for early cancer detection using advanced computer vision on medical imaging. The goal is to deliver a research-grade prototype that detects and prioritizes clinically relevant findings and integrates with standardized reporting workflows, paving the way for a prospective clinical study.

## Role overview

The Junior Researcher will work on-site and implement an end-to-end pipeline: data curation, preprocessing/DICOM handling, anatomical segmentation, classical candidate generation, 3D deep learning–based detection, evaluation, and a lightweight viewer for reader-in-the-loop assessment. Close mentorship and clear milestones are provided.

## Key responsibilities

- Data curation

 - Ingest and organize de-identified imaging studies and associated labels; maintain manifests, provenance, and documentation.

 - Structure labels into clinically meaningful size/risk bins; handle noisy/incomplete annotations.

- Algorithm development

 - Preprocess volumetric CT data (resampling, denoising, normalization; multi-position alignment where applicable).

 - Implement segmentation of relevant anatomy and classical candidate generation to reduce false positives.

 - Train a 3D CNN detector on candidate-centered patches with weak/semi-supervised targets; optimize sensitivity for clinically significant findings while controlling false positives per case.

 - Optional: add a characterization head (e.g., risk/biologic likelihood) where reliable ground truth exists.

- Evaluation and reporting

 - Define robust patient-level splits; compute per-lesion sensitivity by size/risk, false positives per case, and per-case sensitivity.

 - Build simple visualization/overlays and generate standardized, clinically aligned summaries for internal review.

 - Document methods, code, and results; contribute to an internal white paper and potential abstract.

## Must-have qualifications

- B.E./B.Tech/M.Sc./M.Tech in Computer Science, Biomedical Engineering, Data Science, or related fields.

- Hands-on experience with Python, PyTorch/TensorFlow, and medical imaging toolkits (e.g., MONAI, SimpleITK, pydicom).

- Practical knowledge of 3D CNNs/UNet variants, volumetric data pipelines, and GPU training workflows.

- Demonstrated project in medical imaging or volumetric detection/segmentation (GitHub/portfolio or paper).

- Strong experimentation hygiene: Git, reproducible environments, and clear documentation.

- Willingness and ability to work on-site in Kochi, Kerala.

## Nice-to-have

- Experience with CT imaging, anatomical segmentation, or classical computer vision (e.g., curvature-based candidate generation, artifact suppression).

- Familiarity with radiomics and weak/semi-supervised learning for noisy labels.

- Understanding of clinical reporting frameworks and screening metrics.

- Experience building simple viewers (Streamlit/Gradio/ITK widgets) and robust DICOM handling.

- Exposure to data governance for clinical datasets and basic biostatistics.

## What success looks like in 12 weeks

- Curated and versioned imaging subsets ready for training, with label confidence tracking.

- Baseline candidate generator with segmentation and measurable false-positive reduction.

- 3D detection model achieving high sensitivity for clinically significant findings with controlled false positives per case.

- Clear evaluation on a held-out test set and a lightweight reader-in-the-loop demo.

## Tools and stack

- Python, PyTorch, MONAI, SimpleITK/pydicom, NumPy/Pandas  

- Experiment tracking (Weights & Biases/MLflow), Docker/conda, Git  

- Optional: Streamlit/Gradio for viewer, DICOMweb utilities

## What we offer

- Exposure to translational research with potential publication/abstract opportunities.

- Collaborative lab environment, competitive stipend, and performance-based extension.

- Opportunity to impact real-world healthcare workflows.

## How to apply

Email a brief note, CV, and links to relevant projects (GitHub/portfolio/papers) with subject "Junior Researcher – On-site (Kochi) – AI for Early Cancer Detection" to

Include:

- A short paragraph on experience with 3D medical imaging.

- One example of handling noisy or incomplete labels.

- Availability and preferred start date.

- Confirmation of on-site availability in Kochi.

## Application deadline

Rolling review; priority for applications received within 2 weeks.

Note: Prior domain-specific experience is a plus but not mandatory—strong fundamentals, curiosity, and grit matter most.