class ShreyaC: def __init__(self): self.role = "Software Engineer" self.focus = ["Distributed Systems", "Cloud-Native Backend", "Applied GenAI"] self.stack = ["Java", "Spring Boot", "Python", "FastAPI", "Kafka"] self.superpower = "wiring RAG + MCP into real distributed systems" self.education = "MS CS, Santa Clara - AI & Scalable Distributed Systems" def what_i_do(self): return "design event-driven microservices that stay correct under load, " \ "then teach them to reason with AI - without breaking governance."
I build the unglamorous half of software that has to be right: exactly-once payment ledgers, idempotent APIs, zero-downtime key rotation. Lately I bring the same rigor to AI on the backend - governed RAG pipelines, Model Context Protocol tooling, and semantic caching that cuts model cost.
- 🔭 Building secure AI features with Azure OpenAI · LangChain · RAG, connected to enterprise workflows over the Model Context Protocol (MCP)
- ⚡ Scaling event-driven services on Apache Kafka with an eye on p99 latency and throughput
- 🔐 Hardening APIs with Entra ID · OAuth 2.0 · JWT · Key Vault
- 📈 Watching everything through OpenTelemetry · Prometheus · Grafana
- 🌱 Learning deeper vector search (pgvector) and agentic patterns (LangGraph)
Languages & Backend
Java Spring Boot Python FastAPI TypeScript
Distributed Systems & Data
Apache Kafka PostgreSQL MongoDB Redis Apache Airflow
Generative AI
Azure OpenAI LangChain MCP pgvector
Cloud & DevOps
Azure AWS Docker Kubernetes Terraform GitHub Actions
Observability & Security
| Domain | Level | Signal |
|---|---|---|
| Java / Spring Boot Microservices | ██████████ Senior |
3+ yrs, production systems |
| Distributed Systems & Kafka | █████████░ Advanced |
Event-driven, exactly-once, chaos-tested |
| Cloud-Native (Docker / K8s / CI-CD) | █████████░ Advanced |
AKS/EKS, GitHub Actions, Jenkins |
| Applied GenAI (RAG / MCP / pgvector) | ████████░░ Strong |
Shipping AI features at Microsoft |
| Python / FastAPI | ████████░░ Strong |
AI service layer, tooling |
| Observability (OTel / Prom / Grafana) | ████████░░ Strong |
Production monitoring & incident response |
Recruiter-grade, benchmarked, and honestly documented - each solves one expensive problem.
| Project | What it proves | Stack |
|---|---|---|
| llm-guard-gateway | Prompt-injection firewall + semantic cache: p50 44 ms → 4.6 ms | FastAPI · pgvector · MCP |
| txn-exactly-once-ledger | 46 process kills, 0 lost, 0 double-applied transfers | Postgres · Kafka · outbox |
| cdc-read-model-projector | Byte-identical replay; fixes the out-of-order-commit CDC bug | Postgres · CQRS · CDC |
| saga-chaos-lab | 7 crash points, 7 consistent recoveries | Saga · chaos engineering |
📄 distributed-systems papers 🧩 system-design puzzles 🤖 tinkering with AI agents 🛠️ open-source ☕ over good coffee