Founded in 2014 by Harshardhan Lunia, Indian digital assembly fintech lender Lendingkart utilises a data-powered credit analysis system to facilitate online loans, aiming to improve accessibility in small business lending. The company’s proprietary underwriting mechanism utilises big data and...
Founded in 2014 by Harshardhan Lunia, Indian digital assembly fintech lender Lendingkart utilises a data-powered credit analysis system to facilitate online loans, aiming to improve accessibility in small business lending. The company’s proprietary underwriting mechanism utilises big data and analytics to evaluate the creditworthiness of borrowers.
The company has so far disbursed over $1 billion in loans in over 1300 cities in the country, especially in Tier 2 and Tier 3 cities. The company, which recently reported its first-ever profits, a sum of Rs 118 crore, with total revenues reaching Rs 850 crore in FY23, specialises in providing unsecured business loans to micro, small, and medium-sized enterprises (MSMEs).
The fintech company is backed by Bertelsmann India Investments, Darrin Capital Management, Mayfield India, Saama Capital, India Quotient and more.
“Data science has always been at the heart and center of our operations. The AI/ML-based underwriting that this team has developed has been used to underwrite over one million MSMEs,” said Dhanesh Padmanabhan, chief data scientist, Lendingkart, in an exclusive interaction with AIM.
The 35-member data science team of the Ahmedabad headquartered firm is organised into three main groups: analytics, underwriting modelling, and ML engineering. The analytics team, with approximately 15 members, is further divided into three sub-teams focusing on revenue, portfolio (credit and risk), and collections.
“One of the key challenges addressed by our team at Lendingkart is credit risk management where we employ a combination of analytics and AI/ML models at different stages of the underwriting and collections processes to assess eligibility, determine loan amounts and interest rates, and ensure timely customer payments or settlements,” he added.
This underwriting modeling team consists of about 5 members dedicated to developing underwriting models, while the 10-member ML engineering team focuses on MLOps, feature store development, and AI applications.
Additionally, there are individual contributors like an architect and a technical program manager, along with a two-member team specializing in setting up the underwriting stack for the newly established personal loan portfolio.
The company has open positions for senior data scientist and associate director in Bengaluru.
Inside Lendingkart’s AI & Analytics Team
The team leverages AI and ML across various functions, for example, in outbound marketing to target existing customers and historical leads through pre-approved programs. Additionally, a lead prioritization framework helps loan specialists focus on leads for calling and digital engagement.
The company also employs an intelligent routing system to direct loan applications to credit analysts, and a terms gamification framework aids negotiation analysts in negotiating interest rates with borrowers. Its fraud identification framework flags potentially manipulated bank statements for further review, and a speech analytics solution is deployed to extract insights from recorded calls for monitoring operational quality.
On the other hand, collections models prioritize collections based on a customer’s likelihood of entering different delinquency levels, and computer vision models are used for KYC verification.
“We are also exploring the use of generative AI for marketing communication, chatbots, and data-to-insights applications,” said Padmanabhan. Moreover, there are plans to build transformer-based foundational models using call records and structured data sources like credit histories and bank statements for speech analytics, customer profiling, and underwriting purposes.
The tech stack comprises SQL running on Trino, Airflow, and Python. For ML tasks, they leverage scikit-learn, statsmodel, scipy, along with PyTorch and TensorFlow. Natural language processing and computer vision applications involve the use of transformers and CNNs.
The API stack is powered by fast API’s deployed on Kubernetes (k8s). In ML Engineering, the team prefers Kafka and Mongo. Additionally, there are applications built on Flask and Django, and they are currently developing interactive visualizations using the MERN stack.
Interview Process
Lendingkart’s data science hiring process includes four to five interview rounds, evaluating candidates with strong backgrounds in analytics, modelling, or ML engineering. In leadership roles such as team leads and managers, the company places emphasis not only on technical proficiency but also on crucial skills in team and stakeholder management.
During the interview process, non-managerial candidates undergo initial technical assessments in SQL, Python, or ML. Subsequent rounds explore general problem-solving and soft skills, with assessments conducted by peers, managers, and HR.
Expectations
Upon joining the team, candidates can expect to participate in a diverse range of projects encompassing revenue, risk, collections, and the development of tech and AI stacks for these applications. Collaboration with various stakeholders remains a significant aspect of the role. For example, the development of a new underwriting algorithm involves comprehensive reviews with risk and revenue teams to align with business objectives, followed by collaboration with product and ML engineering teams for successful implementation.
However, Padmanabhan notes that there is a common mistake which candidates make – they overlook the importance of thoroughly understanding the business context of the given problems.
“While they may possess knowledge of various algorithms used in different domains, they may struggle to articulate solutions or approaches when those algorithms are applied within a financial process context,” he added, highlighting the importance of connecting technical expertise with a deep understanding of the specific business challenges at hand.
Work Culture
“Our work culture is fast-paced and dynamic, characterised by group problem-solving focused on specific business goals with competitive ESOP packages and industry-standard insurance,” said Padmanabhan.
The data science team operates hands-on at all levels, adopting best practices like agile and MLOps. The “hub and spoke” approach involves data scientists taking responsibility for the entire process from conceptualization to implementation, distinguishing the work culture from competitors in the space.
At Lendingkart, you’ll collaborate closely with stakeholders on projects like developing underwriting algorithms. The company maintains a well-established agile practice led by the technical program manager and team leads, focusing on efficient planning, best practices, and clear communication to create a productive work environment. So if you think you are fit for this role, apply here.
The post Data Science Hiring Process at Lendingkart appeared first on Analytics India Magazine.