An overview of my Machine Learning Internship at Labellerr
About the process and my observations
Introduction
I am Atharva Parikh, a final-year Information Technology student at Vishwakarma Institute of Information Technology, Pune. Before joining Labellerr, I had done two corporate and two research internships, published one research paper in the NLP domain, won the Smart India Hackathon 22, and built multiple projects in the data domain.
My resume at that time can be seen here.
I joined Labellerr on January 23, 2023, with the help of a platform called Cuvette.
Interview Process
Labellerr’s interview process was a well-structured and comprehensive experience, consisting of three rounds aimed at evaluating different aspects of the applicants. Each round played a crucial role in assessing the candidate’s suitability for the ML internship position.
The first round was an informal session with Puneet Jindal, the CEO of Labellerr. This initial interaction served as a getting-to-know-each-other round, where general and scenario-based questions were asked. It became evident that the purpose of this round was to understand the applicant’s nature, behavior, and potential fit within the company culture. The fact that an icebreaker round preceded the technical evaluation indicated Labellerr’s emphasis on building rapport with candidates and valuing those who establish a good connection with the team.
The second round was a task-based technical interview that delved into the specifics of machine learning. In this round, I was assigned the task of training a model to detect potholes in images. Although this task may appear straightforward at first glance, the interviewers explored the real-time implementation aspects and intricacies associated with it. The technical questions primarily focused on assessing the candidate’s understanding of computer vision and object detection fundamentals. The interviewers posed scenario-based questions to gauge the applicant’s problem-solving abilities and knowledge of the challenges involved in deploying a pothole detection system. Some of the questions revolved around the necessity of an automated pothole detection system, the appropriate evaluation metrics (accuracy, precision, or F1 score) for real-time use cases, and the potential challenges that may arise when scaling the system. I have written about the assigned task in detail in the following blog
The third and final round of the interview process involved a discussion with Sumit Singh, the CBO at Labellerr. This round focused on aligning the expectations of both the company and the applicant. Various topics were covered, including the specific tasks expected during the internship, the research interests and goals of the applicant, and the compensation for the internship duration. This round provided an opportunity for open communication and ensured that both parties had a clear understanding of each other’s expectations.
In my opinion, Labellerr’s interview process demonstrated a holistic approach to selecting ML interns. By incorporating a getting-to-know-each-other round, a task-based technical interview, and a discussion on expectations and compensation, the company effectively assesses the candidates’ technical skills, problem-solving abilities, cultural fit, and mutual alignment of goals.
The First Month
In the first month of my internship at Labellerr, my main focus was to explore various autoML tools available on the market. This activity helped me gain a better understanding of the different features offered by each software tool. Additionally, I had the opportunity to use different platforms, which provided valuable user experiences. This exploration allowed me to grasp how different competitive tools present their products, and as a new user of Labellerr’s software, I was able to offer suggestions from a fresh perspective. These suggestions improved the intuitiveness of the product and sparked discussions about potential features for future releases, such as dataset explainability.
I was genuinely impressed by the discipline and high quality of work maintained at Labellerr. Every discussion, code implementation, and thought process involved in developing functionality was meticulously documented. Although I had hoped for more research-oriented work during my internship, I found the experience of diving into data engineering and the MLOps domain to be highly educational and intriguing.
The most challenging aspect for me in the first month was understanding how all the different components fit together and how my individual tasks contributed to the end goal of the product. Moreover, I initially struggled with adapting to the structured working method employed by the company. However, through discussions with my coworkers and managers, I gained a clear understanding of the rationale behind this structure and its benefits for asynchronous working. By “async working,” I mean that all progress is documented with detailed thought processes in JIRA, eliminating the need to constantly inquire about the status of coworkers’ work.
Reflecting on my expectations at the beginning of the internship, I can say that while I initially hoped for more research-oriented work, I am genuinely enthusiastic about the knowledge I have gained in data engineering and the MLOps domain.
My Tasks and Learning Opportunities
- Explored diverse datasets from multiple subdomains in computer vision.
- Conducted an analysis of competitive products and their features.
Engaged in various data engineering tasks. - Explored state-of-the-art research, including DINO, CLIP, ImageBind, and Yolov8, to leverage their potential for Labellerr’s products.
- Trained, tested, and integrated models for object detection, classification, and segmentation.
- Attended and conducted technical presentations on topics such as 3D deep learning, computer vision applications like drowsiness detection, assistive labeling, and the need for MLOps tools like TFX.
- Proofread articles for the marketing team, ensuring technical depth and effective content delivery.
My contributions to the company and my experience are described in detail in the following blog.
Overcoming Initial Discomfort
Initially, I felt quite uncomfortable asking for help and had a strong desire to figure things out on my own. I had a fear of appearing incompetent if I had to seek assistance. As a result, I would spend a considerable amount of time searching for resources and trying to solve problems independently. However, I soon realized that by seeking help, I could save a significant amount of time on exploration, as others had already gone through similar challenges and had valuable insights to share.
For example, I had limited experience with using Postman for API testing, and I had never worked with the AWS cloud before. With the support and guidance of my coworkers, I was able to quickly grasp complex concepts and effectively integrate them into my work. This enabled me to focus more on improving project outcomes rather than struggling with the intricacies of integration.
As I started reaching out for assistance, I discovered that asking for help not only saved time but also fostered collaboration and knowledge sharing within the team. It was refreshing to see how my colleagues were open and willing to provide guidance, which greatly accelerated my learning process.
In hindsight, I realize that my initial reluctance to ask for help was unfounded. It is a natural part of the learning journey, and seeking guidance from others who have more experience is not a sign of incompetence but rather a practical way to enhance productivity and overcome challenges more efficiently.
Suggestions for Future Interns:
- Be prepared for a diverse range of tasks and projects, and remain adaptable to changing priorities and requirements. Labellerr values versatility and agility, so being flexible will allow you to excel in different situations.
- Embrace and welcome changes with an open mind. Labellerr thrives on innovation and continuously seeks to improve and evolve. Embracing change will not only demonstrate your adaptability but also enable you to contribute to the company’s growth and success.
- Actively participate in discussions and seize opportunities to contribute your ideas and insights. Your perspective as an intern is valuable, and sharing your thoughts can spark innovation and new approaches. Don’t hesitate to speak up and make your voice heard.
- Prior to presenting your ideas to the team, consider creating proofs of concept (POCs) to showcase the feasibility and potential impact of your proposals. POCs can provide tangible evidence and help others understand your vision more effectively.
- Demonstrate punctuality and active engagement during meetings. Meetings are not only opportunities to share progress but also valuable learning opportunities and avenues for collaboration. By actively participating, asking questions, and sharing your thoughts, you can deepen your understanding and build stronger connections with your team members.
Labellerr’s Future and Opportunities
Labellerr is on an exciting trajectory, with numerous ML projects and initiatives in the pipeline. The company is continuously pushing the boundaries and aims to launch cutting-edge automl products in the near future.
One notable area of focus is the advancement of autolabel functionality. Labellerr is diligently working towards perfecting this feature, allowing businesses to effortlessly attach their datasets to the platform and have them automatically annotated. This development will greatly streamline and expedite the annotation process for users.
Looking ahead, future interns at Labellerr can anticipate engaging with the latest research and advancements in the field. Whether it’s computer vision, natural language processing, or other relevant areas, Labellerr remains committed to staying at the forefront of innovation. Interns will have the opportunity to contribute to and work on these state-of-the-art projects, gaining exposure to the latest techniques and technologies.
Moreover, the company is dedicated to fostering the growth and development of its interns. Alongside working on groundbreaking projects, interns can expect to strengthen their foundations in data science, programming, and system design. By optimizing the services provided by the platform, interns will contribute to enhancing the overall functionality and efficiency of the company’s offerings.
Conclusion
If you aspire to pursue a career in machine learning, I highly recommend applying for internships at Labellerr. My internship experience has provided me with invaluable learning opportunities, personal growth, and a strong foundation for my professional journey.
Labellerr offers an excellent platform for learning, growth, and kickstarting your career. The diverse range of tasks, the opportunity to work with cutting-edge technologies and datasets, and the supportive and collaborative team environment make it an ideal choice for aspiring machine learning professionals.
