Venkata Manideep Patibandla

I build and ship production AI systems — agents, automated workflows, and reliability layers that solve real engineering problems and hold up under production load.

I built RelayOps, a production-shaped telecom support agent with scoped MCP-style tools, hybrid RAG with citations, guardrails that block invented offers/PII, adversarial agent evals, a Qwen LoRA intent classifier, and a live Railway demo.

I built Tether, a durable execution layer for long-running LLM agents. Wrap your existing OpenAI or Anthropic client and get automatic checkpoint/resume, cross-provider failover, and resilient multi-step agent workflows — without rewriting your agent logic.

I built CostGuard, a production-grade open-source LLM reliability proxy that intercepts calls in real time, auto-rejects low-quality responses via a fallback chain, and tracks exact per-call cost across 12 models and 5 providers — with circuit breakers, Prometheus metrics, and a 6-type alerting engine over Slack and webhooks. Cuts LLM spend 10–20× vs GPT-4o defaults.

To power CostGuard’s model-selection engine, I built RealDataAgentBench (RDAB) — an automated benchmarking system across 12 frontier LLMs and 1,180+ runs on real-world tasks that produces cost-performance trade-off data for production decisions. Key output: GPT-4.1 delivers 97% of GPT-5’s score at 1/15th the cost.

Currently an AI/ML Engineer at Infosoft Solutions, building production ML pipelines, deploying LangGraph and CrewAI agent workflows, and monitoring model quality at scale. IBM Certified in Agentic AI.