About the role

KreditBee is building an AI platform that lets every team ship Gen AI features safely and at scale retrieval, agent workflows, model serving, evaluation, and guardrails. As a regulated consumer-lending business, we hold reliability, consistency, and auditability above all else; this platform is the foundation the org depends on to become AI-native.

It's a single role family with two tracks and a shared engineering bar. You apply to Mandate Skills the family; we place you on the track that fits your depth and the level that fits your experience, and you can move between tracks as the platform grows.

- Software / Platform track — reliable, well-designed distributed systems: the serving, orchestration, data, and tooling layers that make AI usable in production.
- ML / Applied-AI track — making models work: retrieval quality, prompting and agent design, evaluation, and fine-tuning, with production-grade code.
- Owns features end-to-end with light oversight; reliable, independent execution.

What you'll do — shared across both tracks

- Own work end-to-end — design, build, test, ship, monitor, iterate — with production ownership and on-call.
- Write clean, well-tested, maintainable code and clear design docs; apply SOLID and sound API design.
- Treat reliability, latency, cost, and observability as first-class requirements, not afterthoughts.
- Build for a regulated environment: data privacy, access controls, auditability, safe handling of customer data.
- Collaborate across product, data, risk, and platform to turn ambiguous problems into measurable outcomes.

Track A — Software / Platform Engineer

- Build and operate model-serving, gateway,



and orchestration infra (routing, caching, rate-limiting, fallbacks) for LLM/ML workloads.
- Design data and RAG pipelines — ingestion, chunking, embedding jobs, vector/index stores — that stay fresh and consistent.
- Build guardrails, evaluation harnesses, prompt/version management, and observability (tracing, metrics, cost attribution); harden for scale and failure.
- We look for: production backend/distributed systems in a strong language (Go, Java, Python); solid concurrency, APIs, databases, queues, and cloud infra; a reliability mindset. Deep ML theory not required.

Track B — ML / Applied-AI Engineer

- Improve retrieval and RAG quality — chunking, embeddings, re-ranking, grounding — measured against real metrics.
- Build agent and prompt workflows; systematically evaluate models, prompts, and pipelines with offline and online evals.
- Fine-tune, adapt, or distill models where it clearly beats prompting; partner with the platform track to productionize to the same reliability bar.
- We look for: production-quality Python and service ownership (not just notebooks); hands-on LLMs, embeddings/retrieval, and evaluation, plus one of fine-tuning, RAG, or agent frameworks; rigor with data and experiments.

Common bar & nice-to-haves

- Solid CS fundamentals (data structures, algorithms, system design) and a track record of shipping in production.
- Clear communication and a bias for reliability and correctness — especially important in fintech.
- Nice to have: fintech / lending / payments or other regulated domains; LLMOps / MLOps tooling, vector DBs, eval frameworks; open-source or AI/ML side projects.

📌 Software Engineer (Bengaluru)
🏢 Krazy Bee Services
📍 Bengaluru