Software Developer AI - AIGent

BMW TechWorks India Private Limited; Pune District; On-site; INR 1,500,000 - 2,100,000; Full time

Job summary

BMW TechWorks India Private Limited in Pune (Maharashtra) seeks a Software Developer – AI to advance our AI-driven test automation framework for automotive IVI systems. The role blends hands-on coding with framework design and mentorship in a hybrid work setup.

You will design agents, integrate LangChain-like workflows, and ensure robust observability and evaluation pipelines, collaborating with senior engineers to deliver production-grade AI capabilities.

Qualifications

• Strong professional Python experience (production grade code, tests, CI/CD, packaging).
• Solid software architecture skills; design maintainable, modular systems.
• Comfortable with APIs, async code; integrate 3rd party services and tools.
• Practical experience building multi-step or multi-agent systems with LangChain, LangGraph, Semantic Kernel, AutoGen/AG2, CrewAI or equivalents.
• Solid understanding of tool/function calling, structured output, and agent-to-tool interfaces.
• Experience with prompt engineering: versioning prompts, testing, mitigating hallucinations and failures.
• Knowledge of context management techniques (chunking, RAG, memory strategies).
• Experience with LLM observability tools (LangSmith, Langfuse, OpenTelemetry tracing, Arize).
• Experience designing/evaluating LLM/agent output quality (LLM-as-judge, datasets, regression suites).
• Awareness of cost/latency trade-offs between models and selecting appropriate models.
• Ability to read provider docs and adapt systems (OpenAI, Azure OpenAI, Anthropic).

Responsibilities

• Own and improve the overall system architecture (agent orchestration, tool integration, extension).
• Design, build and review new agents and capabilities within the framework.
• Diagnose and fix bugs across the agent pipeline.
• Introduce and maintain observability for agentic workflows (tracing agent, tool calls, token, failure modes).
• Build and maintain evaluation pipelines.

Skills

Python

Software architecture

APIs & Async

LangChain

Prompt engineering

LLM observability

Multi-agent systems

CI/CD

Testing

Tools

LangChain

LangGraph

Semantic Kernel

AutoGen/AG2

CrewAI

OpenTelemetry

Job description

B.Tech/B.E

Job Description

Job Description: Software Developer – AI

Experience: 5+ Years

Role Overview

We are seeking a visionary to help withour next-generation Automotive Infotainment (IVI) systems test automation framework using high capability AI driven approach. You will also be part of our AI-driven test automation strategy, integrating LLMs and Agentic AI workflows to design automation frameworks.

This is a hands-on hybrid role requiring individual coding contribution, framework design review, cross-functional stakeholder management, and technical mentorship for a growing engineering team.

Responsibilities

• Own and improve the overall system architecture (agent orchestration, tool integration, extension)
• Design, build and review new agents and capabilities within the framework
• Diagnose and fix bugs across the agent pipeline
• Introduce and maintain observability for agentic workflows (tracing agent, tool calls, token, failure modes)
• Build and maintain evaluation pipelines

Required skills

• Strong professional Python experience (production grade code, tests, CI/CD, packaging)
• Solid software architecture skills, be able to design maintainable and modular systems
• Comfortable to work with APIs, async code, integrate 3rd party services and tools and collaborate with various external teams
• Practical experience building multi-step or multi-agent systems with a framework such as LangChain, LangGraph, Semantic Kernel, AutoGen/AG2, CrewAI, or an equivalent custom orchestration layer
• Solid understanding of tool/function calling, structured output, and how to design reliable agent-to-tool interfaces
• Experience with prompt engineering: versioning prompts, testing them systematically, understanding failure modes (hallucination, tool misuse, context loss) and mitigating them
• Working knowledge of context management techniques (chunking, RAG, memory strategies) and their trade-offs - not necessarily building embeddings/vector infra from scratch, but knowing when and how to use it
• Experience setting up or using LLM observability tooling (e.g. LangSmith, Langfuse, OpenTelemetry-based tracing, Arize, or in-house equivalents) to debug agent runs in production
• Experience designing or running evaluation frameworks for LLM/agent output quality (e.g. LLM-as-judge, golden datasets, regression suites, automated scoring)
• Awareness of cost/latency trade-offs between models and how to design systems that use the right model for the right step
• Comfortable reading model provider docs/API changes (OpenAI, Azure OpenAI, Anthropic, etc.) and adapting the system accordingly