Posted:5 days ago| Platform: Linkedin logo

Apply

Work Mode

On-site

Job Type

Full Time

Job Description

What You’ll Work On

1.Deep Learning & Computer Vision


  • Train models for image classification: binary/multi-class using CNNs, EfficientNet, or custom backbones.
  • Implement object detection using YOLOv5, Faster R-CNN, SSD; tune NMS and anchor boxes for medical contexts.
  • Work with semantic segmentation models (UNet, DeepLabV3+) for region-level diagnostics (e.g., cell, lesion, or nucleus boundaries).
  • Apply instance segmentation (e.g., Mask R-CNN) for microscopy image cell separation.
  • Use super-resolution and denoising networks (SRCNN, Real-ESRGAN) to enhance low-quality inputs.
  • Develop temporal comparison pipelines for changes across image sequences (e.g., disease progression).
  • Leverage data augmentation libraries (Albumentations, imgaug) for low-data domains.

2. Vision-Language Models (VLMs)

  • Fine-tune CLIP, BLIP, LLaVA, GPT-4V to generate explanations, labels, or descriptions from images.
  • Build image captioning models (Show-Attend-Tell, Transformer-based) using paired datasets.
  • Train or use VQA pipelines for image-question-answer triples.
  • Align text and image embeddings with contrastive loss (InfoNCE), cosine similarity, or projection heads.
  • Design prompt-based pipelines for zero-shot visual understanding.
  • Evaluate using metrics like BLEU, CIDEr, SPICE, Recall@K, etc.


3. Model Training, Evaluation & Interpretation

  • Use PyTorch (core), with support from HuggingFace, torchvision, timm, Lightning.
  • Track model performance with TensorBoard, Weights & Biases, MLflow.
  • Implement cross-validation, early stopping, LR schedulers, warm restarts.
  • Visualize model internals using GradCAM, SHAP, Attention rollout, etc.
  • Evaluate metrics:

• Classification: Accuracy, ROC-AUC, F1

• Segmentation: IoU, Dice Coefficient

• Detection: mAP

• Captioning/VQA: BLEU, METEOR


4. Optimization & Deployment

  • Convert models to ONNX, TorchScript, or TFLite for portable inference.
  • Apply quantization-aware training, post-training quantization, and pruning.
  • Optimize for low-power inference using TensorRT or OpenVINO.
  • Build multi-threaded or asynchronous pipelines for batched inference.


5. Edge & Real-Time Systems

  • Deploy models on Jetson Nano/Xavier, Coral TPU.
  • Handle real-time camera inputs using OpenCV, GStreamer and apply streaming inference.
  • Handle multiple camera/image feeds for simultaneous diagnostics.


6. Regulatory-Ready AI Development

  • Maintain model lineage, performance logs, and validation trails for 21 CFR Part 11 and ISO 13485 readiness.
  • Contribute to validation reports, IQ/OQ/PQ, and reproducibility documentation.
  • Write SOPs and datasheets to support clinical validation of AI components.


7. DevOps, CI/CD & MLOps

  • Use Azure Boards + DevOps Pipelines (YAML) to:
  • Track sprints

• Assign tasks

• Maintain epics & user stories

• Trigger auto-validation pipelines (lint, unit tests, inference validation) on code push

• Integrate MLflow or custom logs for model lifecycle tracking.

• Use GitHub Actions for cross-platform model validation across environments.


8. Bonus Skills (Preferred but Not Mandatory)

  • Experience in microscopy or pathology data (TIFF, NDPI, DICOM formats).
  • Knowledge of OCR + CV hybrid pipelines for slide/dataset annotation.
  • Experience with streamlit, Gradio, or Flask for AI UX prototyping.
  • Understanding of active learning or semi-supervised learning in low-label settings.
  • Exposure to research publishing, IP filing, or open-source contributions.


9. Required Background

  • 4–6 years in applied deep learning (post academia)
  • Strong foundation in:

Python + PyTorch

CV workflows (classification, detection, segmentation)

Transformer architectures & attention

VLMs or multimodal learning

Bachelor’s or Master’s degree in CS, AI, EE, Biomedical Engg, or related field


10. How to Apply

Send the following to info@sciverse.co.in

Subject: Application – AI Research Engineer (4–8 Yrs, CV + VLM)


Include:

• Your updated CV

• GitHub / Portfolio

• Short write-up on a model or pipeline you built and why you’re proud of it


OR apply directly via LinkedIn — but email applications get faster visibility.


Let’s build AI that sees, understands, and impacts lives.

Mock Interview

Practice Video Interview with JobPe AI

Start DevOps Interview
cta

Start Your Job Search Today

Browse through a variety of job opportunities tailored to your skills and preferences. Filter by location, experience, salary, and more to find your perfect fit.

Job Application AI Bot

Job Application AI Bot

Apply to 20+ Portals in one click

Download Now

Download the Mobile App

Instantly access job listings, apply easily, and track applications.

coding practice

Enhance Your Python Skills

Practice Python coding challenges to boost your skills

Start Practicing Python Now

RecommendedJobs for You

Hyderabad, Telangana, India

Bengaluru, Karnataka, India

Kochi, Bengaluru, Thiruvananthapuram