Machine Learning Research Assistant

Deadline: March 25, 2019

Do you want to work on an original research in a multi-disciplinary team that could empower everyone to be able to measure pesticide concentration present in the vegetables and fruits they buy by simply using their smartphones and a filter paper ? See project details below.

Kathmandu Institute of Applied Sciences (KIAS) and Nepal Applied Mathematics and Informatics Institute for Research (NAAMII) announce a vacancy for a position of research assistant to work with Dr. Basant Giri and Dr. Bishesh Khanal.

About KIAS and NAAMII:

KIAS is a not-for-profit research organization in Nepal established by a group of Nepali scientists with proven excellence in basic and applied research as well experience in science popularization. KIAS is an equal opportunity employer dedicated to building a broadly diverse and inclusive researchers and staffs. It believes in promoting affirmative action to achieve such diversities. KIAS encourages all qualified applicants, including minorities and women.

NAAMII is a newly established research institute with a vision of doing world class research to solve local and global problems and help create an ecosystem of high-quality research, innovation and higher education in Nepal. NAAMII is an Equal Opportunity Employer and is committed to diversity and inclusion in our workforce. Women, Indigenous people, marginalised people and people of multicultural backgrounds are encouraged to apply.

Salary:

Rs 30,000 per month

Duration:

Minimum of three months (extensible depending on interest, performance of the candidate and availability of funding**

Application:

Application or any queries must be sent to jobs@naamii.org.np with a cover letter and CV using KIAS-NAAMII-ML0001 as a subject.

Duties and Responsibilities

  1. Do literature review on existing methods

  2. Do research and experiment with an objective to develop better and more accurate method than the current state-of-the-art methods.

  3. Learn about image formation models and computational approaches to inverse problem.

  4. Design experiments and datasets to apply traditional machine learning methods and explore novel deep learning methods to solve inverse problems.

  5. Write research paper based on the work done for publication.

Qualifications

Minimum Bachelor’s Degree in any of the computing sciences or mathematics related field with computer programming experience (Computer/Electrical/Electronics/ Engineering; Mathematics/Physics but knowledge of computer programming; Information Management/Computer Applications or related fields)

Essential and Desired Skills

  1. Willingness to learn and contribute

  2. Good experience with any computer programming language

  3. Desirable: Python, scipy, numpy, scikit-learn (need to learn if not already proficient)

  4. Desirable: Machine learning and mathematical skills (need to learn if not already proficient)

  5. Desirable: Git (need to learn if not already proficient)

Deadline

25 March, 2019

Project details

Paper-based analytical devices (PADs) are new class of devices/methods/tests that are gaining popularity in clinical diagnostics, environmental pollution monitoring and other applications. The PADs are considered to be low-cost, easy to use at point of need and are also praised for not requiring highly skilled personnel to use [1].

Most of the PADs use colorimetric detection method. In colorimetric detection method, the target analyte is allowed to react with a reagent, after they react a characteristic color for the given assay is produced. Just visualizing the color by naked eye may give qualitative result. For quantitative result, photo of the assay color is taken using a camera phone (more commonly a smartphone) and the photo is processed with software such as ImageJ and Adobe Photoshop. Generally, the photo is analyzed on the basis of RGB or HSB/V color space. As such analysis can only give signal of an assay in broad color range, colorimetric paper assays are considered to be less sensitive when compared to their conventional spectrophotometer based assays. In spectrophotometric measurements, signal of a color at given wavelength can be measured, while in paper method signal can be measured as whether it is red or green or blue. Therefore the later method is less sensitive and less selective as it may contain signal from neighboring wavelengths of light.

The image formed by the camera averages out the various wavelengths lying in R, G and B channels. Thus, providing a more sensitive and selective measurement from the images obtained from normal phone camera is an inverse problem. One way to tackle this could be to develop together a computer vision algorithm and the paper analytic device to provide a particular pattern or colors. We are exploring computer vision and machine learning algorithm together with the device development. The computer vision and machine learning algorithm will be explored together with Dr. Bishesh Khanal at NAAMII, while the analytic device development is currently being done at KIAS led by Dr. Basant Giri. The successful candidate will work closely with both the teams to provide a novel framework of device arrangement and suitable machine learning approach to provide accurate measurement PAD.

[1] Sharma, Niraj, Toni Barstis, and Basant Giri. "Advances in paper-analytical methods for pharmaceutical analysis." European Journal of Pharmaceutical Sciences 111 (2018): 46-56. full-text-PDF