Shishir Kumar

At Agilo Inc., I build shared AI agent orchestration platforms and MCP-based infrastructure that let engineering teams develop, deploy, and operate production AI applications at scale. I launched a production-grade AI platform and agent framework within three months, supporting multi-agent workflows, tool calling, agent memory, state persistence, and multiple LLM providers.

I design FastAPI services, LLM guardrails, authentication and authorization for agent-tool interactions, MCP server templates, and centralized tool registries. I also build evaluation, monitoring, and feedback systems using MLflow, DeepEval, Grafana, distributed tracing, and GitHub Actions to improve safety, reliability, latency, and cost performance.

Previously at Larus Technologies, I developed MLOps frameworks for NLP classification models and built an LLM-powered FastAPI application for generating military scenarios integrated with simulation tools. I also delivered Python services that classified and enriched 50k+ enterprise documents with BERTopic and transformer-based NLP models.

My earlier data engineering and software engineering work included Airflow, PySpark, CI/CD, data mappings, ETL automation, Python, VBA, and Unix scripting. I led open-source automation adoption at Tata Consultancy Services, trained a team of five, and helped achieve 30% cost savings.