Hello there, I'm

Aakaash Panigrahi.
I build scalable AI solutions.

Machine Learning Engineer passionate about building AI solutions in deep learning, computer vision, natural language processing, and generative AI.

View My Work Get In Touch
Aakaash Panigrahi

About Me

I'm Aakaash Panigrahi, a Backend Developer and Machine Learning Engineer passionate about solving real-world challenges with AI. I specialize in deep learning, computer vision, NLP, and generative AI, leveraging cutting-edge technologies to build impactful solutions.

My professional experience includes developing and optimizing machine learning models for various applications. I've worked on high-performance AI models, data-driven systems, and backend infrastructure, ensuring seamless integration of AI solutions in production environments.

Beyond technical expertise, I have a strong interest in AI research, model optimization, and large-scale data processing. My projects involve building deep learning models, optimizing machine learning pipelines, and developing backend systems to support AI-driven applications.

I'm always eager to collaborate on exciting AI-driven projects that merge innovation with real-world impact. Let's connect and explore possibilities together!

3
Years Experience
3
Projects Completed
3
Papers Published
10+
Technologies Mastered

My Skills

I specialize in deep learning and generative AI, focusing on designing neural architectures, integrating them into production workflows, and leveraging generative models for real-world AI solutions.

Programming Languages

  • Python
  • Java
  • SQL
  • Bash

Backend Development

  • Flask
  • FastAPI
  • AirFlow
  • MySQL
  • MongoDB

AI & Data Science

  • NumPy
  • Pandas
  • Agno
  • PyTorch
  • TensorFlow
  • HuggingFace

DevOps

  • Docker
  • Kubernetes
  • CI/CD Pipelines
  • Linux Systems

Featured Projects

Here are some of my recent projects showcasing my skills in Computer Vision, NLP and Generative AI.

Gen AI & NLP
Ongoing
NanoBots

NanoBots

Lightweight, chainable LLM agents designed for workflow automation and infrastructure interoperability

NanoBots is a modular suite of lightweight LLM-based agents built to automate everyday workflows across diverse domains. Designed with scalability and composability in mind, each agent is independently deployable and can be chained to form more sophisticated AI systems for task orchestration.

Built on top of modern agentic frameworks, NanoBots emphasizes performance, interpretability, and low-resource execution—making it ideal for both personal use cases and enterprise-grade automation. It serves as a flexible foundation for integrating AI agents into existing backend systems, enhancing productivity through intelligent decision-making and streamlined execution.

  • Composable and modular agent architecture with LLM backbone
  • Supports low-latency, low-footprint deployments for edge or cloud
  • Designed for plug-and-play integration into automation pipelines
Python Agno LLMs RAG EmbedChain Streamlit
CV & Backend
Completed
GeoEngine

GeoEngine

Bridging research and production in remote sensing and geospatial ML

GeoEngine is an end-to-end platform designed to support the development, training, and deployment of geospatial AI models at scale. It enables researchers and practitioners to build production-ready ML pipelines for satellite imagery analysis with robust data handling, visualization, and evaluation workflows.

  • Supports full ML pipeline from data ingestion to deployment
  • Modular architecture enabling reproducible experimentation and benchmarking
  • Used in research accepted at CVPR 2022 for geospatial AI applications
Python PyTorch FastAPI AirFlow Docker
CV
Completed
MRI Style Transfer

MRI-Transform

Contrast-to-contrast MRI transformation using unpaired image translation

This project focused on improving diagnostic flexibility in medical imaging by training a CycleGAN model to convert T1-weighted MRI scans into T2-weighted scans—and vice versa—without needing paired datasets. The model preserved structural integrity across contrasts, achieving a Structural Similarity Index (SSIM) of 97%, validating the effectiveness of contrast transformation in enhancing radiological interpretation and multi-modal training.

Built using TensorFlow, the pipeline was designed to support seamless model training, evaluation, and visualization using Matplotlib. The final models were used to generate synthetic contrast images that mimic clinical scans, helping augment datasets where certain MRI modalities are underrepresented.

  • Trained CycleGAN on unpaired T1 and T2 MRI images for bi-directional style transfer
  • Achieved 97% SSIM, ensuring high structural fidelity in generated outputs
  • Enabled synthetic data generation to boost training for downstream medical tasks
Python TensorFlow CycleGAN Medical Imaging Matplotlib

More Projects Coming Soon

Check back later for more exciting projects!

View All GitHub Projects
See All Projects

My Publications

Here are my research contributions in geospatial intelligence. My research has focused on developing datasets, platforms, and methodologies to advance geospatial research and improve satellite imagery analysis.

AI & ML
Published
Publication Image

Have Foundation Models Seen Satellite Images?

Investigating the Generalization of Pre-Trained Models on Geospatial Data

This study evaluates the zero-shot performance of pre-trained foundation models on remote sensing tasks. The research analyzes whether these models have previously encountered satellite imagery and assesses their adaptability to standard benchmarks like EuroSAT and BigEarthNet-S2. Additionally, it explores the impact of geospatial domain-specific textual descriptions compared to standard class-based prompts.

  • Assesses foundation models' zero-shot classification abilities in remote sensing.
  • Compares standard class-based prompts with geospatial domain-specific descriptions.
  • Evaluates model performance on EuroSAT and BigEarthNet-S2 datasets.
Zero-Shot Learning Foundation Models Remote Sensing
Geospatial AI
Published
Publication Image

GeoEngine: A Platform for Geospatial Research

Bridging Research and Production in Geospatial AI

GeoEngine is an end-to-end geospatial AI platform that enables researchers to seamlessly transition from research prototypes to production-ready geospatial applications.

  • Provides an AI-powered framework for geospatial data processing.
  • Integrates deep learning models for remote sensing and mapping.
  • Designed to bridge academic research with scalable real-world use cases.
Geospatial AI Remote Sensing Machine Learning
Change Detection
Published
Publication Image

QFabric: A Multi-Task Change Detection Dataset

Advancing AI for Urban and Environmental Change Analysis

QFabric introduces a multi-task geospatial dataset designed for change detection, improving AI model performance in tracking urban growth, deforestation, and climate-related changes.

  • Curates high-resolution satellite images for multi-task change detection.
  • Supports tasks like segmentation, classification, and object detection.
  • Aims to standardize change detection benchmarks in geospatial AI.
Change Detection Geospatial AI Computer Vision
See All Publications

Get In Touch

I'm currently open to job opportunities and collaboration on interesting projects. Whether you have a question or just want to say hi, I'll do my best to get back to you!

Say Hello
Email Call LinkedIn GitHub