AI Engineer (Autonomous Agents, Voice and MCP)

PraajX Private Limited; Pune District; On-site; INR 4,000,000 - 8,000,000; Full time

Job summary

SWAB Technologies Private Limited, based in Pune, seeks an experienced AI engineer to build an autonomous AI agent from the ground up. You will own the end-to-end development, from planning and tool integration to deployment and long-term system ownership.

You will craft real-time voice pipelines, design MCP servers, enable actions like calls and ad campaigns, and implement guardrails and monitoring to ensure quality and security.

Qualifications

• Significant hands-on experience building and shipping production software.
• End-to-end capability: backend, APIs, integrations, deployment without heavy specs.
• Live product, repository, or demo you built and can explain.

Responsibilities

• Build an autonomous AI agent that understands voice commands and acts on its own.
• Design MCP servers and tool integrations for real business tools.
• Enable the agent to place calls, create ad campaigns, and perform tasks.
• Build real-time voice pipelines with low latency for Indian accents/languages.
• Develop the agent core: planning, tool calling, tasks, retries, memory.
• Add guardrails to demand confirmations for irreversible actions.
• Create evaluations, monitoring, and cost/latency controls.
• Translate founder requirements into working software without detailed specs.
• Own the system architecture, reliability, security, and documentation.

Skills

LLM agent

Voice product

Python

Node.js/TypeScript

OAuth

Security

Tools

MCP protocols

Job description

SWAB Technologies Private Limited is a Pune-based technology company. We are building an autonomous AI agent that takes voice commands and carries out real business work on its own, such as making calls and running ads.

Role description

We need an experienced AI engineer. This is not a role for freshers or for someone learning on the job. You will build this agent from the ground up. The founder defines what it must do. You decide how to build it, you build it end to end, and you own it for the long term.

• Build an autonomous AI agent that understands voice commands and takes action on its own
• Design and build MCP servers and tool integrations so the agent can operate real business tools
• Enable the agent to act: place and handle phone calls, create and run ad campaigns, and carry out other everyday business tasks
• Build realtime voice pipelines (speech-to-text, LLM, text-to-speech) with low latency, supporting Indian accents and languages
• Build the agent core: planning, tool calling, background task execution, retries, and memory
• Add guardrails so the agent asks for confirmation before irreversible actions such as spending ad budget, payments, or calls, and defend against prompt injection
• Build evaluations, monitoring, and cost and latency controls, so quality is measured and not assumed
• Turn the founder's requirements into working software without needing detailed specifications
• Own the system long term: architecture, reliability, security, and documentation

Company description

SWAB Technologies Private Limited is a Pune-based technology company. We are building an autonomous AI agent that takes voice commands and carries out real business work on its own, such as making calls and running ads.

Role description

We need an experienced AI engineer. This is not a role for freshers or for someone learning on the job. You will build this agent from the ground up. The founder defines what it must do. You decide how to build it, you build it end to end, and you own it for the long term.

• Build an autonomous AI agent that understands voice commands and takes action on its own
• Design and build MCP servers and tool integrations so the agent can operate real business tools
• Enable the agent to act: place and handle phone calls, create and run ad campaigns, and carry out other everyday business tasks
• Build realtime voice pipelines (speech-to-text, LLM, text-to-speech) with low latency, supporting Indian accents and languages
• Build the agent core: planning, tool calling, background task execution, retries, and memory
• Add guardrails so the agent asks for confirmation before irreversible actions such as spending ad budget, payments, or calls, and defend against prompt injection
• Build evaluations, monitoring, and cost and latency controls, so quality is measured and not assumed
• Turn the founder's requirements into working software without needing detailed specifications
• Own the system long term: architecture, reliability, security, and documentation

Qualifications

• Experienced engineer required: significant hands-on experience building and shipping production software, including at least one LLM-based agent or voice product used by real users
• Able to build alone, end to end: backend, APIs, integrations, and deployment, without a large team or detailed specs
• Experience with tool calling, structured outputs, asynchronous jobs, and MCP or similar tool-integration protocols
• Strong in Python or Node.js/TypeScript
• Experience with realtime voice or telephony, or a working project that shows you can learn it quickly
• Understands security for connected business accounts: OAuth, token storage, and least-privilege access
• Based in Pune, or ready to relocate to Pune for this on-site role
• Can show a live product, repository, or demo you built and explain every major decision

Nice to have:

• Experience with advertising platform APIs such as Meta or Google Ads
• Telephony or SIP integration experience
• Indian-language speech recognition or text-to-speech
• Early-stage startup experience as one of the first engineers

What we value:

• Ownership: treats the product as theirs, not as a ticket queue
• Self-driven: takes a clear goal and delivers without being managed step by step
• Long-term commitment: this is a multi-year build, and we want someone who plans to stay
• Direct and honest: tells the founder when an idea will not work, with reasons
• Pragmatic: ships a working version, then improves it with real usage