First seen by Alion on Nov 16, 2026.
JOB DESCRIPTION :
Role : Applied AI Engineer II
Blitz is hiring a hands-on Applied AI Engineer II : an individual contributor who ships production AI agents end to end. You will build many agents across customer-facing and internal workflows, and own each from prototype to measured production performance.
What you'll build :
Blitz runs 60-minute to same-day delivery for leading consumer brands. We want AI agents to absorb repeatable work across the company, so the business grows faster than its headcount. This is a builder role, not a research role.
- Conversational agents : chat and voice agents that resolve real user queries in production, at scale.
- Workflow agents : agents that take on multi-step operational tasks, call tools and systems, and know when to hand off to a human.
- Evaluation systems : the eval sets, graders and monitoring that tell us whether an agent is working and getting better.
- Shared agent platform : reusable harnesses, tool integrations, context and memory patterns that make each new agent faster to ship than the last.
What you'll do :
- Own agents end to end : scope the problem with the team that has it, design the prompts, tools and context, ship to production, and keep improving it.
- Ship fast and in parallel. Expect to take a new agent from idea to live pilot in weeks, and to run several at once.
- Treat evals as core craft : define success metrics, build eval sets from production data, and gate every prompt or model change on them.
- Monitor agents after launch : read transcripts and traces, find failure patterns, and fix them at the root.
- Integrate agents with backend systems and data through clean, well-scoped tool interfaces.
- Track cost, latency and quality per agent, and make the model and architecture choices that improve them.
- Adopt new technology early : when a new model, framework, protocol or tool lands, try it within days, judge it on our evals, and put what works into production.
- Build what does not exist yet : when no off-the-shelf tool fits, design and ship the harness, integration or internal tool yourself, and get the rest of the team using it.
- Work directly with business teams. There is no PM layer between you and the user.
What we're looking for :
- 2 - 5 years of software engineering, with at least one LLM agent shipped to real users and run in production. Demos and thin wrappers do not count.
- Strong Python, and solid backend fundamentals : APIs, queues, databases, cloud deployment.
- Hands-on agent craft : prompt and context design, tool calling, retrieval, multi-step workflows, and handling non-deterministic failures.
- Eval discipline : you have built eval sets, used LLM-as-judge or human review, and can show how a change moved a metric.
- Heavy daily use of AI coding tools to move at several times normal speed.
- A track record of picking up new technology fast : you were early on tools your team later adopted, and you are not attached to last year's stack.
- Comfort with ambiguity : you pick up a loosely defined problem and return with a working agent, not a design doc.
- Clear written communication and ease working with non-technical teams.
Nice to have :
- Voice agent experience : speech-to-text, text-to-speech, telephony, latency tuning.
- Work with multilingual or Indic-language models.
- Fine-tuning or serving open models for cost and latency.
- Experience building in a high-volume operations or consumer business.
Stack and how we work :
- Models : frontier LLM APIs, with open models where cost and latency demand it.
- Agent tooling : Model Context Protocol (MCP) for tools, an LLM observability stack for tracing and evals.
- Infra : AWS, with standard CI/CD.
- Team : a small applied AI group, reporting to the CTO.
- Compliance : SOC 2 Type II, so data handling is part of the design, not an afterthought.
- Location : Bengaluru, in office.
Skills
Artificial Intelligence, Machine Learning, Python, Generative AI, LLM, Agentic AI, Conversational AI

