Founding AI engineer

InCommon; Bengaluru; On-site; INR 1,500,000 - 2,100,000; Full time

Job summary

InCommon, a seed-stage consumer fintech building an AI-first personal finance app for India, is hiring to own the intelligence layer of the product. You’ll design and implement the prompting layer, evaluation infrastructure, and the orchestration of multiple agents that route tasks to specialist tools.

You will work in a founding-team seat, shipping early AI systems, wiring the knowledge base to the user context, and ensuring reliability as the system scales with new capabilities.

Qualifications

• 3–6 years building applied AI products, ideally in a zero-to-one setting.
• Hands-on prompt engineering experience with shipped evals.
• Experience setting up agent harnesses; strong grasp of tool calling.

Responsibilities

• Build and iterate on the prompting layer of the core answering engine.
• Set up evaluation infrastructure to move fast on prompt and model changes without regressing quality.
• Design and build the agent orchestration layer - routing, multi-agent pipelines, and handoffs between specialist agents.
• Work on tool calling: what to expose, how to handle failure modes, how to manage context when multiple tool calls are in flight.
• Connect the financial knowledge base to the user context layer through the orchestration layer.

Skills

Applied AI product development

Prompt engineering

Agent orchestration

Tool calling

Context management

Infra exposure

Tools

EC2

Job description

InCommon is hiring on behalf of a seed-stage consumer fintech building an AI-first personal finance app for India - a premium, conversational alternative to the dashboard products that dominate the category. The thesis is that most people don't deploy their money well, not because the information is missing, but because it's buried in charts that assume you already know what to look for. The product reads a user's full financial context and tells them what matters, proactively.

You’ll own the intelligence layer of the product. Corpus works by combining what it knows about a user with a knowledge base of financial information, routed through a set of specialist agents - one for mutual fund advisory, one for tax, one for stocks, and more over time. Your job is to make that system actually work, and keep it working as it gets more complex.

This is a founding-team seat. You're not tuning an existing pipeline - you're building the one the company runs on.

What you’ll do

• Build and iterate on the prompting layer of the core answering engine
• Set up evaluation infrastructure so the team can move fast on prompt and model changes without regressing quality
• Design and build the agent orchestration layer - routing, multi-agent pipelines, and handoffs between specialist agents
• Work on tool calling: what to expose, how to handle failure modes, how to manage context when multiple tool calls are in flight
• Connect the financial knowledge base to the user context layer through the orchestration layer

What we’re looking for

• 3 - 6 years building applied AI products, ideally in a zero-to-one setting
• Real, hands-on prompt engineering experience - shipped and iterated against evals, not just experimented in a notebook
• Experience setting up agent harnesses; strong grasp of tool calling and where it breaks
• Understanding of context management in multi-agent systems
• Comfortable in an unstructured environment with no playbook
• Bonus: deployment and infra exposure (EC2, keeping services running)