Krutrim, founded by Ola's Bhavish Aggarwal, is building India's sovereign AI stack: LLMs trained on Indian languages, inference infrastructure, and AI-native products.
The engineering team is small and moves fast, so interviews are thorough but lean at 3-4 rounds. Expect deep questions on ML systems, LLM inference optimisation (batching, quantisation, KV-cache), and distributed training infrastructure if you are applying for AI infra roles. For application engineering roles, the focus shifts to API design, low-latency services, and integrating LLM capabilities into products. Krutrim is an early-stage unicorn; interviewers value candidates who can operate with ambiguity and build from scratch.
About Krutrim
Krutrim (meaning 'artificial' in Sanskrit) is India's first AI unicorn, founded by Bhavish Aggarwal (Ola's founder) as a spinoff. It builds Indian-language AI models and is working on custom AI chips.
Apply via Krutrim careers page or LinkedIn; referrals carry significant weight at this stage
Initial screening call with a recruiter or engineer (30 minutes)
Technical rounds: coding problem-solving and system design or ML systems discussion
Founder or senior-leader round: vision alignment, ability to work with ambiguity, first-principles thinking
Offer and compensation negotiation
Screening Call (30-45 min)
background discussion, motivation for joining Krutrim, and one light technical question.
Technical Round 1 (60 min)
1-2 coding problems at medium-hard difficulty; may include system design for an LLM-adjacent feature.
Technical Round 2 (60 min)
deep-dive on a past project or an ML systems design question covering inference, batching, or training infrastructure.
Founder or Senior-Leader Round (45 min)
strategic thinking, first-principles problem decomposition, and culture fit for a high-velocity startup.
Sourced from 2+ candidate post-mortems. Hit Practice to answer any one with AI voice feedback.
2 more questions. Sign up to unlock all
Sign up free: unlock all questionsThe typical Krutrim recruitment process has 5 stages: Apply via Krutrim careers page or LinkedIn; referrals carry significant weight at this stage → Initial screening call with a recruiter or engineer (30 minutes) → Technical rounds: coding problem-solving and system design or ML systems discussion → Founder or senior-leader round: vision alignment, ability to work with ambiguity, first-principles thinking → Offer and compensation negotiation.
Krutrim typically conducts 4 interview rounds: Screening Call (30-45 min): background discussion, motivation for joining Krutrim, and one light technical question.; Technical Round 1 (60 min): 1-2 coding problems at medium-hard difficulty; may include system design for an LLM-adjacent feature.; Technical Round 2 (60 min): deep-dive on a past project or an ML systems design question covering inference, batching, or training infrastructure.; Founder or Senior-Leader Round (45 min): strategic thinking, first-principles problem decomposition, and culture fit for a high-velocity startup..
HireStepX recommends the AI-infra design framework for this type of interview: Model serving latency budget → batching strategy → quantisation trade-offs → KV-cache design → auto-scaling → cost per token.
To answer this question well, HireStepX recommends the AI-infra design approach: Model serving latency budget → batching strategy → quantisation trade-offs → KV-cache design → auto-scaling → cost per token. Ground your answer in a specific real example from your own experience.
To answer this question well, HireStepX recommends the AI-infra design approach: Model serving latency budget → batching strategy → quantisation trade-offs → KV-cache design → auto-scaling → cost per token. Ground your answer in a specific real example from your own experience.
To answer this question well, HireStepX recommends the AI-infra design approach: Model serving latency budget → batching strategy → quantisation trade-offs → KV-cache design → auto-scaling → cost per token. Ground your answer in a specific real example from your own experience.
To answer this question well, HireStepX recommends the AI-infra design approach: Model serving latency budget → batching strategy → quantisation trade-offs → KV-cache design → auto-scaling → cost per token. Ground your answer in a specific real example from your own experience.
To answer this question well, HireStepX recommends the AI-infra design approach: Model serving latency budget → batching strategy → quantisation trade-offs → KV-cache design → auto-scaling → cost per token. Ground your answer in a specific real example from your own experience.