Computer Vision - Internship

We are looking for an AI Engineer with a deep understanding of neural networks

and other AI techniques, as well as traditional image processing techniques,

capable of reasoning from first principles. The role demands the ability to

analyze the data and problem statement, select appropriate features, model

architectures, loss functions, and evaluation metrics, and design or modify

neural networks accordingly.

The ideal candidate is not a "model-user", but a model designer and problem

solver, comfortable working with limited, noisy, or domain-specific data, and

combining classical computer vision with modern deep learning approaches.

Selected intern's day-to-day responsibilites include:

A. Problem Analysis & Model Selection:

1\. Analyze problem statements, operational constraints, and available data to

determine:

2\. Suitable feature representations

3\. Supervised vs unsupervised vs self-supervised vs traditional image

processing approaches

4\. Appropriate network architectures and learning paradigms, algorithms

5\. Select and justify loss functions, regularization strategies, and

evaluation metrics based on task objectives.

B. Design and Development:

1\. Design neural network architectures from scratch for:

2\. Classification, detection, similarity matching, retrieval, etc.

3\. Modify and adapt existing architectures (CNNs, Siamese networks,

autoencoders, transformers, etc.) for domain-specific requirements.

4\. Optimize model size, latency, and accuracy for deployment constraints.

C. Classical Computer Vision and Hybrid Approaches:

1\. Apply and integrate traditional image processing and vision techniques

such as:

2\. Bag of Visual Words (BoVW)

3\. Feature descriptors (SIFT, SURF, ORB, HOG, etc.)

4\. Image similarity, matching, and retrieval

5\. Develop hybrid pipelines combining classical vision and deep learning

where appropriate.

D. Training, Evaluation and Optimization:

1\. Design training pipelines, data augmentation strategies, and validation

methodologies.

2\. Diagnose training issues such as overfitting, underfitting, class

imbalance, and convergence problems.

3\. Perform error analysis and iterate on feature sets, architectures, and

losses.

E. Collaboration and Documentation:

1\. Work closely with domain experts, systems engineers, and software teams.

2\. Document model choices, assumptions, and design trade-offs clearly.

3\. Support integration of models into larger software or embedded systems.

What We Bring

Work on operational defense and strategic UAV systems delivering real impact.

Contribute to the next generation of autonomous aerial navigation technologies

for India's defense ecosystem.

A high-autonomy, innovation-driven environment with hands-on access to flight

systems.

Opportunity to collaborate with leading research organizations, DRDO labs, and

global technology partners.