Backend & Distributed Systems Engineer | Kafka · Spark · Java · Python · AWS
Building agentic AI systems in production. Building in public on LinkedIn.
An open-source agentic AI system that monitors real-time Kafka streams, detects anomalies autonomously, and either fixes them or escalates to a human — powered by 6 specialized AI agents.
- 🔍 WatcherAgent — detects anomalies on live Kafka streams
- 🧠 DiagnosisAgent — local Llama 3.2 via Ollama explains root cause in plain English
- ⚖️ BlastRadiusAgent — deterministic BFS graph traversal scores downstream impact (zero AI — governance decisions must be auditable)
- 🔧 RemediationAgent — auto-fixes LOW/MEDIUM, escalates HIGH to human
- 📰 NarratorAgent — auto-generates incident post-mortem report
- 🧠 MemoryAgent — pgvector stores every incident as embedding, retrieves similar past incidents as context
Stack: Python · Kafka · Ollama · Llama 3.2 · pgvector · PostgreSQL · Prometheus · Grafana · Docker
View Project · Follow the build on LinkedIn
- 🖥️ Software Development Engineer @ Capital Group
- 🔭 Building StreamSentinel — agentic AI on top of Kafka streams
- 🌊 5+ years building production data pipelines at scale
- ⚡ Passionate about distributed systems, real-time data, and AI agents
- 📍 Based in California, United States
Agentic AI pipeline intelligence for real-time financial data streams
| Feature | Details |
|---|---|
| 🔍 Detection | Real-time anomaly detection on live Kafka streams |
| 🧠 Diagnosis | Local Llama 3.2 — no data leaves the machine |
| ⚖️ Blast Radius | Deterministic BFS — LOW / MEDIUM / HIGH impact scoring |
| 🔧 Remediation | Auto-fixes LOW/MEDIUM, pages human for HIGH |
| 📰 Narration | Auto-generates plain-English incident post-mortem |
| 🧠 Memory | pgvector stores past incidents, retrieves similar context |
| ❤️ Health | Health check endpoint + Prometheus metrics + Grafana dashboard |
🔗 github.com/Neelam95/StreamSentinel
⭐ Currently building StreamSentinel in public — follow along on LinkedIn!