Generative AI Engineer

Job description

**Why we are hiring an AI Pod?** AI is changing how lending works. The opportunities for Kissht sit across five capability areas: - Document intelligence across Indian languages - Voice AI in Indian languages - Agentic workflows - Internal AI productivity - Measurement and evaluation We are setting up a dedicated AI Pod to drive work across these capability areas — and to set the template for how AI work gets done across Kissht. **Responsibilities — building and delivery** - AI features end-to-end across Voice, LAP, Customer Service, and Onboarding — from prototype to production, depending on what each initiative needs at the moment. - Internal tools that make the pod faster: eval dashboards, prompt playgrounds, data explorers, integration scaffolding. - Production integrations between external AI vendors and Kissht systems — APIs, webhooks, data pipelines, observability. - Compliance with Kissht’s data and security standards. You design for PII safety from the first commit. - Pair work with the AI Product Lead and Tech Lead to translate specs into shipped artefacts. **What success looks like — first 90 days** - You have shipped working code on at least two of the four initiatives. - At least one thing you built is in daily use — either by the pod (an internal tool) or by a business team (a production feature). - You have a working relationship with at least three Kissht engineering teams whose systems the pod touches. **What success looks like — first 180 days** - You are the named owner of at least one AI feature in production. - You have moved fluidly between greenfield prototype work and production integration work, depending on initiative need — not stayed inside one lane. - You have started embedding more directly with one business team — owning AI outcomes inside their roadmap, not just inside the pod's. **Must-haves** - 3–5 years of engineering experience. Strong Python. Comfortable with backend services, REST, async patterns, and webhooks. - Genuine versatility. You have shipped quick prototypes and you have run production integrations. You do not specialise in only one of those. - High comfort with LLM APIs. You have built non-trivial things on production-grade LLMs. - Bias for action. You decide and move with incomplete information, ship a working thing in 2–3 days when needed, and stand behind it in production. - Proof of building. GitHub repos with real code are required. Side projects, hackathon work, or open-source contributions count. A polished LinkedIn alone is not enough. **Nice-to-haves** - Frontend skills (React or similar) — useful for quick demos and internal tools. - Experience integrating with telephony providers or contact-centre platforms. - AWS production experience, familiarity with Streamlit/Gradio/Chainlit, and feature flag tooling. - Snowflake experience or comfort working against a cloud data warehouse.