Website:
tellm.co
Job details:
Company Description
At Curious, we're building AI-native consumer products for Indian households - products that understand how people actually live, shop and run their days.
Agents are at the core of what we build: they read messy real-world data, reason over it and take action on behalf of users.
Based in Bengaluru, we're a small team building what we hope becomes the default way Indian households manage themselves.
Role Description
We're looking for an AI Engineer to own the agentic layer of our product - from the harness and orchestration up to the evals that tell us whether things are actually working.
You'll be hands-on with LangGraph every day, designing multi-step agents that run in production against real user data.
Unlike a typical big-company role, you'll be involved in everything: prompt design, graph topology, tool integrations, evals, tracing, cost/latency tuning and the occasional gnarly debugging session when a model starts behaving differently than it did yesterday.
What you will do:
- Design and ship multi-step agent systems in LangGraph for real consumer use cases.
- Build and evolve the agent harness: orchestration, state management, tool use, memory, retries, interrupts, human-in-the-loop, checkpointing.
- Set up evals and observability (LangSmith or similar) so we actually know whether changes are improvements, regressions or noise.
- Make judgment calls across the reliability / cost / latency / capability tradeoff space, and revisit them as models and prices shift.
- Work with product to scope what's possible with today's models and what to hold back for better ones.
- Get hands-on with prompt design, model selection, structured output and fine-tuning where it earns its keep.
- Integrate agents with structured and unstructured data sources (order history, emails, APIs) while keeping context windows honest.
- Debug the failure modes nobody has written a blog post about yet.
What you should bring:
- 4–6 years shipping software, with meaningful recent time on LLM and agent systems.
- Hands-on LangGraph experience: multi-step agents, complex state machines, tool orchestration, checkpointing, streaming.
- Solid grasp of agent best practices: prompt engineering, evals, guardrails, tracing, cost and latency management.
- Strong software fundamentals - good agents are software engineering first, ML second.
- Fluency with Python and the modern AI stack (Anthropic / OpenAI APIs, vector stores, LangSmith or equivalent).
- First principles thinker - the field shifts weekly, and you're comfortable throwing away last month's approach when something better appears.
- High agency, AI-native workflow, creative problem solver.
- Comfort in an early-stage environment: incomplete specs, changing priorities and the need to make progress with limited resources.
Nice to have:
- Consumer-facing AI product experience where latency, cost and UX actually matter.
- Experience with RAG, fine-tuning, or structured data extraction at scale.
- Background in ML / NLP or exposure to applied research.
How we work:
Small, tight-knit team where product, design and engineering sit together and make decisions quickly.
You'll own big pieces of the product and see your decisions reach real users within weeks, not quarters.
We care about solid engineering over polished decks - if you're happiest when evals, prompts and agent graphs are moving forward, you'll fit right in.
Bengaluru, in-office.
Click on Apply to know more.