Talk to us

Role: AI & Data Engineering Intern

Employment Type: Intern

Educational Qualification: Currently pursuing or recently completed a Bachelor's degree in Computer Science, Machine Learning, Data Science, Engineering, Physics, or a related field.

Role Description:

We are seeking an AI & Data Engineering intern to support Pixxel’s participation in NTRO’s AI based change detection competition - a final-stage, high-stakes technical challenge sponsored by India’s national remote sensing agency. This is a time-bounded, intense internship (Competition duration: roughly 8-12 weeks) where you’ll work across the full stack of our competition deliverables: from data pipelines and annotation infrastructure, through AI model development, to evaluation dashboards and agentic

Interfaces.

Responsibilities & Duties:

Data,Pipelines&Infrastructure:

• Designed and built training and benchmarking datasets across two satellite sensors and eight change detection tasks, establishing consistent annotation standards and quality control processes.

• Built and optimized scalable data pipelines for scene ingestion, preprocessing, chip generation, and dataset export, enabling efficient handling of multi-sensor satellite data.

• Developed an interactive dashboard to evaluate and demonstrate change detection results for competition judges, providing clear visualizations of detected changes, confidence scores, and model performance metrics.

• Optimized model inference performance by improving latency, throughput, and memory utilization to meet competition runtime and resource constraints.

AI & Machine Learning

• Develop a change captioning system that generates natural language descriptions of changes by leveraging two time-step satellite images and a change detection mask. This may involve vision-language models, multimodal embeddings, or custom deep learning architectures.

• Build the agentic interface layer for interacting with change detection outputs, enabling users to query, interpret, and reason about detected changes through conversational or programmatic interfaces.

• Develop a sensor-agnostic super-resolution model for satellite imagery that supports multiple upscaling factors (2×, 4×, and 8×). Evaluate trade-offs between model complexity, accuracy, and inference performance.

• Benchmark and validate models against ground truth datasets and competition evaluation metrics. Continuously iterate on model architectures and training strategies based on performance feedback.

Cross-Functional Responsibilities

• Write clean, efficient, and maintainable Python code for data processing and model development. Contribute to shared utilities and machine learning libraries used across the competition team.

• Collaborate closely with cross-functional teams by:

• Providing feedback on dataset quality and annotation issues.

• Partnering with the ML team on model evaluation and performance improvements.

• Coordinating with data engineers to optimize data pipelines.

• Document datasets, models, pipelines, and technical processes comprehensively to ensure reproducibility and provide clear technical insights for team members and competition judges.

Desirable Skills & Certifications:

• Strong programming skills in Python with hands-on experience using libraries and frameworks such as PyTorch, TensorFlow, NumPy, Pandas, or equivalent.

• Solid understanding of deep learning fundamentals, including Convolutional Neural Networks (CNNs), attention mechanisms, loss functions, optimization techniques, and evaluation metrics.

• Practical experience or strong academic coursework in data pipelines, data processing, or data engineering workflows.

• Basic understanding of the end-to-end machine learning lifecycle, including model training, evaluation, and deployment at scale.

• Ability to work independently on well-defined workstreams, take ownership of deliverables, and proactively seek guidance from senior scientists when needed.

• Strong attention to detail, with an understanding that high-quality data is critical to model performance.

• Eagerness to learn quickly, adapt to new challenges, and thrive in a fast-paced, competition-driven environment.

Candidate Acumen:
You will own end-to-end workstreams across two primary areas: (1) Data & Infrastructure — building training datasets (with consistent annotation standards across two satellite sensors and eight change detection tasks), constructing data pipelines, and optimizing inference; and (2) AI/ML — developing change captioning systems (vision-language models that describe detected changes in natural language), building super-resolution capabilities (sensor-agnostic up scaling at 2x, 4x, 8x), and creating the agentic interface or interactive change detection. This is a hands-on role that bridges data engineering and deeplearning –you’ll work closely with senior scientists and ML engineers on the competition team.

Benefits:
Health insurance coverage
Unlimited leaves & flexible working hours
Role-based remote work and work-from-home benefit
Relocation assistance
Professional Mental Wellness services
Creche facility for primary caregivers (limited to India)
Employee Stock Options for all hires

About Pixxel

Pixxel is a space data company and spacecraft manufacturer redefining Earth observation with hyperspectral imaging. The company’s first three commercial hyperspectral satellites—Fireflies—deliver imagery at 5-meter resolution and 135+ spectral bands, providing 50x richer detail than traditional Earth observation systems and unlocking insights across agriculture, climate, energy, environment, and more.

Once fully deployed, Pixxel’s constellation of 18-24 satellites will capture imagery across up to 250 bands in VNIR and SWIR ranges, with a 40 km swath and daily global revisit capability. Pixxel’s most unique strength is its full-stack approach, integrating every layer of the value chain from satellite hardware and manufacturing to AI-powered analytics.

Pixxel’s satellite constellation is complemented by Aurora, its in-house Earth Observation Studio that simplifies satellite imagery analysis and democratises remote sensing for all. Designed to make hyperspectral data more accessible, Aurora by Pixxel combines high-frequency imagery with AI-powered tools to generate actionable insights, even for users without technical backgrounds. The third pillar of Pixxel’s ecosystem is its in-house satellite manufacturing capability. Beyond building its own spacecraft, Pixxel also provides satellite systems and subsystems to other organisations. This dual capacity sets it apart in a sector where most companies focus on either payload design or data operations, but not both.

Pixxel’s team is young but deeply mission-aligned, with a culture rooted in curiosity, speed, and long-term thinking. As the company grows its constellation and expands Aurora, the focus remains on making space-based insights practical, scalable, and genuinely helpful so that the health of the planet becomes measurable and action becomes possible.

Pixxel was the only Indian startup selected for the Techstars Starburst Space Accelerator in Los Angeles and has been recognised in TIME’s Best Inventions of 2023, Fast Company’s Most Innovative Companies, and Via Satellite’s Top Innovators list.

For more information, see our Careers page or follow Pixxel on LinkedIn and X/Twitter.

Culture

Pixxel is an organization where we enable our employees to work on world-changing problems that they are passionate about, and a place where they can be their best selves day in and day out while ensuring they have fun every day.

Central to our employee-first ethos is a commitment to ensuring that every team member feels valued and heard. We nurture a healthy and supportive work environment where we prioritise well-being and growth in all aspects of life
Link to Pixxel in Media