Chingari is one of India’s fastest growing short-video apps. The platform attracts close to 35+ million monthly active users (MAU) and has recorded close to 110+ million app downloads. By the end of 2022, it is planning to grow its user base to 200+ million. 

The company is now looking at various initiatives to increase user engagement across the country and plans to increase its local content portfolio by deploying 20+ local languages on the platform. Chingari has been investing heavily in AI and analytics, where it is working on various AI/ML models (recommendation engines) to offer personalised experiences to its users. 

But, the question is, how is Chingari different from other TikTok clones? To this, Chingari’s CTO, Tariq Wali, said that they are taking a holistic view of content, users, and creators to build a comprehensive end-to-end workflow that solves the enrichment of metadata, content moderation, and recommendation from video upload to video delivery in real-time. 

Wali said that Chingari is a content company; we translate massive amounts of data (textual, video, and audio). The engagement and retention of users depend on how this data is leveraged and therefore making the right decisions on what content is shown to users across languages and various user preferences. 

“While users primarily come to watch the video that’s churned out from our content recommendation system, the predictions made by the systems are important, and in a way, the recommendation system itself becomes the product,” said Wali. 

He said that the data science team has a huge impact on the company’s overall success. Besides the recommendation systems, the team also runs a content management system that moderates the user video upload and classifies them to what could be premium content worthy of being surfaced on the feed. 

“The whole user experience of watching the content, liking, sharing, and spending time on the app is tied to AI/ML systems, powering the content in the backend through recommendation system, search powered by NLP as our video uploads at scale, moderated by moderation systems (video classifiers and computer vision algorithms),” shared Wali. 

For instance, the team recently deployed a multi-model recommendation system to address cold start and hot start. For those unaware, an app launch can occur in one of three states, each affecting how long it takes for an app to become visible to the user. This includes cold start, warm start or hot start. In cold start, the app starts from scratch. In hot start, on the other hand, the system needs to bring the running app from the background to the foreground. 

In an expansion mode, Chingari told AIM that it is looking to hire data science professionals across various experiential levels (4-10 years of experience) in Bengaluru, offering an average CTC of INR 20 lakh for the junior positions and above. Currently, it has close to six job openings. 

Team structure 

“We are a humble team of eight and growing,” said Wali. The data science team at Chingari currently includes data engineers, data analysts, ML engineers, data scientists, and NLP experts. The team closely works with product, backend, and content teams. 

Interview process 

Chingari said that there are three-four rounds of interviews, starting with exploratory calls to determine mutual interest, followed by technical interviews (data science and also, coding), sometimes working on an assignment, and lastly, overall fitment in the company. This includes technology, skills/craft check, behaviour, bar raiser, etc.

Skills in focus 

Technology capabilities: Python, PyTorch, TensorFlow, GitHub, MongoDB, and AWS. 

Data science & ML skills: Candidates need to have proficiency in machine learning techniques like classification, regression, and sequential modelling. Plus, they need to have experience in handling large-scale tabular data and image/video data. 

Others: The candidate should have relevant experience building a recommendation system and image and video classification systems. 

In addition to this, deep learning knowledge like neural networks, convolutional networks, transformers, etc., is a plus. 

A big NO: to those candidates who lack understanding of machine learning fundamentals and lack relevant machine learning projects. 

Expectations from candidates 

Wali said that the candidates need to understand big data, AI/ML libraries and frameworks, MLOps, AWS, good problem-solving skills, and coding. Plus, he said that the person should have closely worked with product teach teams and internalised that to AI/ML deliverables. 

He said that candidates should also provide a quality user experience and be scientific about the efficiency and accuracy of ML models. Also, they need to be aware of the latest research/trends in the AI/ML space and take ideas from the latest research papers to solve complex problems innovatively. 

Work culture 

“A great work culture, amazing colleagues, exciting problems,” said Wali. 

Further, he said data science teams in other companies are not intimate with the product and user experience as they operate in silos. “At Chingari, we strive to solve/build what ultimately impacts users, providing great satisfaction,” he added. 

Why join Chingari?

“We operate in a space unlike another tech company in the country, where there is an interesting intersection between AI/ML and blockchain,” said Wali. He said they are the world’s first AI/ML-powered social media company transitioning to the blockchain. 

Wali said they are working on challenging learning-based problems under the big data regime. “Our solutions focus on using current SOTA methods to solve these problems. We provide access to large-scale annotated data and cutting-edge computing so you can focus solely on building the solution,” he added, stating that Chingari is a platform that offers a great opportunity to solve end-to-end problems.  

Click here to apply.