Vishal V.

I build agentic AI systems that run in production - not prototypes. A couple of years ago my work was mostly Elasticsearch and data infrastructure. It's shifted a lot since then. These days I'm building full AI pipelines: multi-agent workflows with LangGraph and CrewAI, RAG systems on Pinecone and Qdrant, voice AI agents that actually make phone calls (Vapi and ElevenLabs), and security automation that goes from alert ingestion to response without waiting for a human. Some of the more interesting things I've shipped: - An automated SOC workflow that triages SIEM/XDR alerts through L1/L2 AI agents and runs response playbooks - no analyst needed for first-response decisions - A full sales pipeline where AI handles outreach, runs live qualification calls, matches buyers to sellers, and closes deals with e-signatures. Humans only touch exception cases - An employee compliance agent that monitors Slack, has actual conversations with flagged employees, and escalates to an L2 agent trained on company policy via RAG - not just a flag-and-alert system - An AI-driven personalized search engine where results rank differently per user based on real behavioral signals, not a shared relevance score My infrastructure background is deep in Elasticsearch and OpenSearch - cluster design, multi-tenant architectures, ingestion pipelines, and performance work. I once took a 50M-document production cluster from visibly slow to sub-second and called it an 8x improvement. I've also built a full multi-tenant CSOC SaaS on AWS OpenSearch with Wazuh handling the security telemetry. Stack I work with day-to-day: Python (FastAPI, Flask), LangChain, LangGraph, CrewAI, n8n, Make, Supabase, Pinecone, Qdrant, Kafka, Docker, Kubernetes, AWS. LLMs: GPT-4o, Claude, Groq, Llama locally. Voice: Vapi, ElevenLabs, Twilio, Voiceflow. I stay until the thing works. Message me if you have a project that needs a real system.