Job Summary
Hybrid Mode
Immediate Joiner Only
Experience:-7-12 Years
Location:- Pune
We are seeking a Performance Engineer with solid experience in AI agents, AI platforms, distributed systems, and performance engineering to design, execute, and optimize performance testing for next-generation AI applications. The ideal candidate will have hands-on experience testing LLM-based applications, AI agent workflows, APIs, microservices, cloud platforms, and high-scale distributed systems.
Key Responsibilities
Design and implement performance, load, stress, scalability, endurance, and capacity testing strategies for AI agents and AI platforms.
Develop performance test plans, workload models, test scenarios, and acceptance criteria for LLM and AI-powered applications.
Test AI agent workflows involving LLMs, RAG, vector databases, tools, APIs, memory, orchestration, and multi-agent systems.
Measure and optimize latency, throughput, concurrency, scalability, resource utilization, availability, and reliability.
Analyze LLM inference latency, token throughput, time-to-first-token (TTFT),



time-per-output-token (TPOT), request latency, and cost/performance metrics.
Build automated performance testing frameworks using tools such as JMeter, Gatling, Locust, k6, or custom Python-based frameworks.
Generate realistic workloads for concurrent users, API requests, AI agent sessions, prompts, tokens, and multi-step agent workflows.
Conduct performance testing of REST APIs, GraphQL APIs, microservices, event-driven systems, and asynchronous workloads.
Profile applications and identify performance bottlenecks across application, database, network, infrastructure, model-serving, and cloud layers.
Monitor and analyze systems using Prometheus, Grafana, ELK/OpenSearch, Datadog, New Relic, CloudWatch, Azure Monitor, or equivalent tools.
Evaluate performance of AI infrastructure including GPU/CPU utilization, memory, network throughput, model serving, autoscaling, and containerized workloads.
C

📌 Performance Tester (Pune)
🏢 Innova ESI
📍 Pune