My experience with the SpaceML mentorship program for CAMS
Take a look up at the night sky. What do you see? Chances are you will spot the odd bird, aircraft, and if you’re lucky, a meteor whizzing past. Similarly, if I took a look at the night sky from a different geographic location, I might make a similar observation.
It is very hard to correlate perspectives of the night sky from different locations. We know that we all see the same moon every night. But how do we ascertain what an object is when it’s the size of a pea held at an arm’s length? Further, for such small objects moving fast across the sky, how do we conclude that observations coincide with the same object? This is only possible through multiple observations from different places.
For decades, scientists have relied on re-evaluations of previous earth observations to correlate that they belonged to the same event. This was mostly due to a lack of information sharing and limited communication methods.
The confirmation of preexisting meteors and identification of new ones is extremely important. Scientists believe that life on Earth might be traced back to microorganisms brought by meteors and comets. Thus understanding the origin of meteors, their patterns of appearance and their recovery is crucial to understanding our very origins. This is why Scientists have advocated for a network to track and study meteor showers. An example of this is Dr Peter Jenniskens and the CAMS network.
CAMS stands for the Cameras for All-Sky Meteor Surveillance. It is a project and a method, led by Dr Peter Jenniskens of the SETI Institute in partnership with SpaceML. It is aimed at scaling up meteor detections in the night sky through the use of multiple observations. CAMS uses a network of cameras to record a range of activities in the night sky — from the peering of cats into cameras, tail lights of planes, the trails of satellites, the paths of migrating birds and of course meteor showers! Footage provided by CAMS is invaluable. For example, have you ever wanted to name a celestial object? Well, Dr Jenniskens did that, using the CAMS network to identify a new meteor shower he named the Arids!
The CAMS network has set up cameras at different locations across the globe, with the aim of setting up a global meteor hunting network. CAMS operates cameras in the United States, South America, Europe, southern Africa, the Middle East and Australia. However, the network has no cameras operating in Asia and there is a genuine gap in earth observations from that part of the earth. In order to form an inescapable net for catching meteors, the CAMS network needs to extend its coverage into the Indian Subcontinent.
My journey with the CAMS project started with the intention of helping to build techniques that would bridge this gap in observations. As I embarked on the Indian extension of the program, I learnt valuable techniques and gained significant experience working with peers internationally, and I’d like to share this journey through this post.
How did I learn about CAMS?
My first exposure to CAMS came when I heard about my institute’s involvement in the project. Swastik Shinde, a member of the CAMS India extension, shared details of the institute’s involvement with the astronomy club WhatsApp group. I consider myself lucky to have heard about this project, given that although I had a keen interest in astronomy, I hadn’t joined the astronomy club group until the beginning of my final year at university. This was mostly due to the limited size and outreach of the club. I was also unsure who to express my interest to. The timing of Swastik’s message was perfect!
The SETI Institute is the stuff of legends. SETI has been at the forefront of searching for life in the universe. The name SETI extends far beyond the confines of the science community, being the subject of several books and movies. When I heard that SETI was involved in the CAMS project, I immediately got in touch with Swastik to express an interest in and learn more about the project.
A couple of weeks after expressing my interest, Swastik informed me that Siddha Ganju, the SpaceML CAMS AI project lead, was looking for students to help her develop open-source software for the project. Though I found the opportunity exciting, I was hesitant to apply. I believed that as a student of basic science, I didn’t have the computer science background that Siddha was seeking. Nevertheless, I decided to give it a shot. This was a chance for me to explore and expand my horizons, in tune with the ethos of the SETI Institute.
I was keen on applying for the software dev role as the project description mentioned pipeline construction. Pipeline construction had been till then a recurring theme in my previous projects. I saw a chance to apply what I had learnt in python coding from previous projects to an area I had little to no exposure to. My reaction to being told I had been selected was one of gratefulness. I was more surprised to hear that I was the only one selected for the job. It was gratifying to learn that someone believed enough in my abilities to give me a chance on a project of the magnitude of the CAMS project.
After being shortlisted, I had a formal interview with Siddha. This interview not only gave her the opportunity to gauge my fit with the project but also allowed me to learn more about CAMS and my role in the project. Following my successful interview, we created a project plan for my work.
While I was originally brought in to optimize the pipeline, Siddha saw an opportunity for me to use my past experiences and also work on constellations for the cams web portal. Thus constellations were the first thing I worked on as a part of the CAMS project, followed by optimizing the data processing pipeline. Read about my work on constellations and data processing.
Balancing CAMS and academics
Contrary to what people might believe, CAMS fit my academic schedule really well. Despite being in the final and most intense year of my graduate studies, CAMS did not interfere with my academic work. CAMS also did not compromise on my leisure time. In fact, CAMS gave me a break from the monotony of school work.
More often than not, I would work on CAMS in the evening, leaving ample time in the morning for me to complete my assignments. Working in the evenings also gave me the chance to interact with and meet my colleagues in various global locations. I think this was one of the best parts of the project, working with a team that worked internationally, and knowing that there was always someone around to support you should you need it. Here is a brief overview of the timeline of my work.
I did not allocate fixed times per day for either academics or CAMS. Rather, depending on the urgency of academic assignments, I would allocate more or less time to CAMS in a day. In a week, I would spend 15–17 hours on the CAMS project. For the first three months, I worked on creating constellations. Towards the end of March, I started work on optimising the CAMS pipeline. Unfortunately towards the end of April, I had a bout of covid. I restarted work on CAMS after finishing my final exams, roughly in June. In September, optimization was complete. Since then, I have been writing this blog, making presentations and writing a paper!
A supportive team
While I had used python a lot before this project, I was unfamiliar with some of the tools needed for the project, such as Unix programming and accessing remote servers. Siddha helped me through a period of initial difficulties and hiccups. She provided me with learning resources, cleared doubts and guided me through certain procedures when I was well and truly stuck. Siddha has been and is extremely approachable, even when doubts seemed silly or mundane. Siddha’s mentorship provided the platform and motivation I needed to make progress on CAMS.
Siddha constantly emphasised the need for continuous testing and meticulous attention to detail. She is an ardent proponent of the scientific method, always encouraging me to come up with hypotheses and create experiments to test code. My work in CAMS involved a lot of experimentation and testing, This is something that translates well into my own work as a student of science. All in all, she introduced me to good research practices and habits, which I have included in my work with SpaceML as well as elsewhere.
Interacting with Siddha has introduced me to the world of AI and ML, and the type of work being carried out by the community. Siddha has been persistent in her encouragement for me to showcase my work. I am by nature an introvert and have always shied away from public speaking or even large scale interaction. She has introduced me to people within the community and has provided avenues/opportunities to showcase my talent and helped improve self-confidence.
All my colleagues — Siddha Ganju, Surya Ambardar, Alfred Emmanuel, Chicheng Ren, Julia Nguyen, Chad Roffey and Meher Anand Kasam, are exceptionally talented. Their love for their work was visible in the quality of the output as a part of the CAMS project. I was keen to interact with them and Dr Jenniskens to get their viewpoint, learn from their expertise, and have some of their enthusiasm rub off on me. My colleagues and I formed a global team where anyone could discuss issues, raise suggestions and implement ideas at any time of the day. We were able to accomplish a lot in limited time and uncertainty in the backdrop of a global pandemic. The CAMS project is a testament to teamwork from oceans apart.
The experience of a lifetime
I believe that CAMS and SpaceML set a precedent for citizen science in AI. Employing mentors who are leaders in their respective fields not only brings exposure to students and laymen, it also allows industry professionals to bring their expertise into applied areas of their work. In return, it provides experts with a platform to showcase their work to a large, distinct audience. Citizen science projects like CAMS also allow researchers like Dr Jenniskens to avail of advancements in the emerging fields of AI and ML. Online mentorship and citizen science allow researchers to tap into a global pool of talent to assist them in their work.
Guidance by experts not only instils confidence in students, it also inculcates responsibility and professional etiquette. This drives students to produce work of greater quality. Mentorship programs provide invaluable experience to students, especially when they move on to conducting their own research. Interdisciplinary research like in CAMS encourages students to develop multiple interests or explore how they can apply their skills to a new domain. This is invaluable in a tumultuous, ever-changing, fast-paced world.
My time with SpaceML has been a massive personality and skill-building exercise. The CAMS project gave me an opportunity to showcase my work at international conferences, global live streaming events, blogs (this one!) and research papers. These experiences have improved my speaking, writing and presentation skills. These skills are valuable to any field requiring the dissemination of information to the public. This is why I believe mentorship programs like SpaceML can train budding professionals possessing a wide range of skillsets.
My time with SpaceML has imparted me a renewed appreciation for code. While working on CAMS, I picked up a professional coder’s mindset and etiquette. I have learned of the importance of readability and robustness of code, all while improving the performance and scalability of code.
I wholeheartedly recommend students join the SpaceML community. For students, SpaceML provides an avenue to learn more about the broad-spectrum application of Machine Learning and AI. At the same time, it provides a platform to make contributions to the betterment of humanity through high quality, purpose-made code. I also recommend SpaceML to researchers as a toolbox to advance research and improve rates of significant discoveries about our planet, solar system and beyond!
In the new world order created by the pandemic, SpaceML and my experience with CAMS showcases a model of how people from different parts of the world with different skill sets, can collaborate successfully. It is evidence that geopolitical borders are no boundaries to impactful work.
