Agentic AI Engineer

Job Title: Agentic AI Engineer

Location: Bangalore (HSR Layout) 

Why This Role Exists

Tazapay is building AI-native infrastructure across payments, compliance, and merchant operations. We believe AI agents are not experimental tools, they are production systems that must operate reliably inside a regulated financial environment.

This role exists to design, deploy, and scale production-grade AI agents that reduce manual effort, improve decision accuracy, and create measurable business impact. 

You will own the internal AI automation roadmap and help define how intelligent systems operate within a fintech company.

What You’ll Own

1️. Build Production-Grade AI Systems

    • Design, deploy, and maintain AI agents that automate workflows across onboarding, compliance, reconciliation, support, and internal operations.

    • Architect systems with logging, monitoring, audit trails, and fallback mechanisms.

    • Ensure AI workflows are safe, explainable, and regulator-ready.

2️. Prompt Architecture & Agent Design

    • Design structured prompt systems with guardrails, schema validation, and deterministic fallbacks.

    • Orchestrate multi-step agent workflows combining LLMs, APIs, and rule-based systems.

    • Balance deterministic components with probabilistic AI reasoning.

We are not looking for prompt hobbyists — we are looking for system designers.

3. Evaluation & Reliability Engineering

    • Build evaluation frameworks to measure accuracy, hallucination rates, and reliability.

    • Design regression tests and offline evaluation datasets.

    • Monitor model drift and failure modes in production.

    • Implement human-in-the-loop controls where necessary.

AI without evaluation is experimentation. We build durable systems.

4️. Cost & Performance Optimization

    • Optimize model selection based on cost, latency, and task complexity.

    • Implement caching, batching, and pre-processing strategies to reduce token usage.

    • Architect workflows that scale economically under high operational load.

5️. Translate Operational Pain → Automation Impact

    • Work closely with product, ops, and engineering teams to identify automation opportunities.

    • Convert ambiguous problems into structured AI workflows with measurable outcomes.

    • Define ROI metrics and track performance over time.

6️. Own the AI Stack & Roadmap

    • Evaluate and integrate LLM APIs, agent frameworks (LangChain, CrewAI, AutoGen, etc.), orchestration platforms (n8n or similar), and MCP-compatible systems.

    • Make pragmatic build vs buy decisions.

    • Document systems clearly so they are extensible and reusable.

What We’re Looking For

    • 3–6 years of experience in a technical, analytical, or product-adjacent role.

    • Demonstrated hands-on experience building and deploying AI agents or LLM-based systems in production.

    • Strong systems thinking ability to design for reliability, scale, and failure recovery.

    • Experience working with APIs, structured data, and workflow automation.

    • Strong analytical mindset ability to measure what works and communicate clearly.

We will ask you to walk us through:

    • What you built

    • Architecture decisions you made

    • What broke

    • How you debugged it

    • How you measured success

    • How you optimized cost

Technical Exposure

Experience with some of:

    • OpenAI / Anthropic APIs

    • LangChain, CrewAI, AutoGen, or similar frameworks

    • MCP-based architectures

    • Workflow orchestration tools (n8n or similar)

    • Logging, monitoring, and observability systems

    • SQL or working with structured datasets

Bonus Points

    • Experience in fintech, payments, or other regulated environments.

    • Background in product management or product analytics.

    • Familiarity with compliance automation, risk scoring, or workflow engines.

    • Experience building evaluation datasets and model benchmarking frameworks.

What Success Looks Like (12 Months)

    • Manual effort reduced by 40%+ across at least 3 major operational workflows.

    • Agent reliability consistently above defined accuracy thresholds.

    • Clear evaluation and regression testing framework in place.

    • AI cost per workflow optimized and tracked.

    • Internal AI automation stack becomes reusable, extensible, and documented.

    • Playbooks created for building safe, production-ready agents at Tazapay.

Why This Role Is Different

You won’t be building demo bots.

You’ll be designing intelligent systems that operate inside a regulated payments company, where reliability, auditability, cost efficiency, and measurable outcomes matter.

If you’ve built serious AI systems and want to apply them to real financial infrastructure problems, this is the role.