CRED AI Engineer Fresher Track: The Credit Card Layer Reality

CRED built three production AI systems in a single year: Cleo answers member queries at 98% accuracy, Thea summarises support conversations, and Stark manages operations knowledge. Freshers who want to work on any of these enter through one realistic route.

What CRED Is and Why the ML Problem Is Different

CRED is a fintech platform built specifically for India’s creditworthy households. It launched in 2018 as a credit card bill payment app, and the premise was unusual: only users with a CIBIL score above 750 could join. That filtering decision shapes everything about the ML problem at CRED.

Most consumer apps work with noisy, heterogeneous user bases. CRED works with a relatively homogeneous, high-credit-score population, which means behavioural signals are denser, fraud patterns are more subtle, and the margin for error on credit intelligence is smaller. A missed fraud signal on a high-value credit card transaction costs more than the same miss on a low-value UPI payment.

As of 2025, CRED serves over 15 million premium members. Spread across bill payments, rewards, credit score monitoring, and curated offers, that scale means ML models at CRED run inference millions of times per day. Fraud detection, credit scoring, personalised rewards ranking, and the AI concierge layer are all live, production-grade ML surfaces. Not research projects.

CRED is incorporated as Dreamplug Technologies Pvt Ltd and is headquartered in Bangalore. All engineering roles are Bangalore-only. The company does not run mass campus drives and does not publish a standard placement paper pattern. Entry for freshers is selective and overwhelmingly through the intern pipeline.

CRED’s AI Layer in Production: Cleo, Thea, Stark, and PROTECT

In late 2025, CRED declared itself an AI-first fintech platform through a formal partnership with OpenAI. Three internal AI tools went into production as a result.

Cleo is the member-facing conversational AI. Built on GPT-4 and GPT-5, it handles three classes of queries: informational (what is my credit limit?), contextual (why was I charged a late fee last month?), and transactional (pay my HDFC bill now). According to the CRED-OpenAI partnership case study, Cleo achieves 98% resolution accuracy across all three classes.

Thea is an internal tool for CRED’s support agents. It summarises multi-format conversations (chat, email, call transcripts) so agents spend less time reading history and more time solving problems.

Stark manages SOP updates for the operations team. When a process changes, Stark propagates the update across relevant documentation rather than relying on manual edits.

Across all three tools, CRED measured a 14-percentage-point CSAT increase, a 31% reduction in session drop-offs, and an 18% improvement in multi-intent resolution success. These figures come from the CRED-OpenAI case study, which is the primary source for all verified outcomes metrics.

The fourth AI surface is CRED PROTECT, the payment security layer. It monitors every step of a credit card payment for fraud in real time. The PCI Security Standards Council featured CRED PROTECT in their AI Exchange series on payment security innovation, an external validation that CRED’s approach to fraud detection is being watched as a reference model, not just a product feature.

Freshers who want to contribute to any of these systems need to understand what problems they’re actually solving: real-time inference under latency constraints, LLM orchestration at scale, fraud signal modelling on high-credit-score populations, and reward-ranking personalisation. These are production engineering problems, not tutorial-scale exercises.

The Three Fresher Tracks at CRED

CRED does not have one universal fresher hiring route. Three tracks exist, and each has a different realism score for Tier-2 and Tier-3 college candidates.

Software Engineer Intern (Primary Fresher Route)

This is the most realistic entry point. CRED runs multiple intern cohorts per year. A January 2026 cohort was listed on careers.cred.club, confirming active intake. The selection process:

• Online coding assessment (DSA focus)
• One or two technical interview rounds (problem-solving and data structures)
• Team matching call
• Stipend estimate: ₹60,000 to ₹1,00,000 per month (aggregator data; CRED does not publish official figures)

Target candidates are penultimate or final-year CSE, IT, or ECE students. Tier-2 and Tier-3 college candidates who clear the coding assessment and perform well in interviews do convert. The intern route is explicitly not limited to IIT/NIT students, unlike the direct SWE track.

Software Engineer Backend or Platform (Direct FTE)

This track flows almost entirely through intern-to-FTE conversion or referrals. Direct campus hiring here skews heavily toward IITs and top NITs. Selection includes DSA rounds and a distributed systems design interview. Core stack is Go, Kotlin, and gRPC microservices. Key details:

• CTC band: ₹22 to ₹30 LPA for entry-level SWE (aggregator estimate; not officially disclosed by CRED)
• All roles based in Bangalore
• No mass campus drive; targeted university partnerships only

Machine Learning Engineer (Research-Background Freshers Only)

CRED’s ML Engineering role formally requires two or more years of experience. Exceptional freshers with NLP or LLM fine-tuning research publications, and at least one notable prior internship, are occasionally considered. The portfolio and GitHub review precedes the technical rounds, which cover ML systems design and algorithm implementation. Key details:

• CTC estimate: ₹30 to ₹60 LPA depending on experience (aggregator data; not officially disclosed)
• Covers active ML surfaces: fraud detection, credit scoring, personalised rewards ranking, and conversational AI
• Bar-raiser behavioural round included in the process

This track is not the first target for most freshers; the intern route is.

What CRED’s Interview Process Looks Like

The intern and SWE interview process at CRED is straightforward compared to mass-recruiting IT services companies. There is no aptitude test, no English proficiency section, no verbal reasoning module. The entire technical process is DSA and systems.

For intern candidates, two rounds is typical. Round one is a coding assessment: expect medium-to-hard LeetCode-calibre problems. Round two is a technical interview that may include problem-solving discussion, a system-level question, and a culture and values conversation.

For SWE Backend FTE roles, the process adds a distributed systems design round. Given CRED’s microservices architecture on Go and gRPC, questions about service communication, consistency, and fault tolerance appear regularly. CRED’s engineering blog publishes case studies on their actual distributed architecture. Reading it before an interview is one of the more direct ways to understand what problems the team is working on.

For ML Engineering, the process includes a portfolio review, a ML systems design round (covering model deployment, feature pipelines, and latency tradeoffs), and a coding round focused on ML algorithm implementation. A bar-raiser behavioural round follows.

One thing the process does not include: a formal CGPA cutoff listed in any public job posting. In practice, IIT and top NIT candidates dominate direct hiring partly because of referral density, not necessarily because of a published filter. For Tier-2 candidates, the intern route avoids that screen entirely.

Building the Skills CRED’s ML Layer Is Hiring For

CRED’s AI stack does things that are genuinely teachable through small projects. Cleo’s resolution accuracy benchmark, cited at 98% in the CRED-OpenAI case study, is built on the same class of LLM orchestration that any engineering student can prototype. CRED PROTECT’s real-time fraud signals use the same feature-engineering and inference patterns covered in standard ML coursework.

The gap between knowing those patterns and demonstrating them on a live project is what separates intern candidates who receive offers from those who don’t. CRED’s ML interview starts with a portfolio review, which means the project must exist before the application, not after.

For a wider view of how AI skills map across 2026 hiring decisions beyond CRED, the AI roadmap for 2026 covers the full curriculum sequence and how companies across sectors are interpreting “AI-skilled” differently. For context on how other companies with similar AI-readiness filters are structured, the breakdown of AI readiness for campus freshers is a useful comparison.

The fastest way to ship a real LLM project without getting stuck on API setup is TinkerLLM. At ₹299, it gives access to live LLM API calls (the same class of GPT inference that powers Cleo), and the resulting micro-project is what goes on the resume the next time a recruiter asks what you’ve actually built. The difference between a CRED intern application that says “familiar with LLMs” and one that links to a working conversational AI prototype is the difference that gets a response.