Raj
Mandaliya
Specializing in systems programming, distributed infrastructure, and production AI engineering — building high-performance Rust services, event-driven Kafka pipelines, and end-to-end ML systems at cloud scale.
Experience
- Designed and implemented high-performance backend services using Rust, handling 10M+ requests/day with <50ms p95 latency and comprehensive performance analysis.
- Built async microservices using Tokio and Actix Web, improving system throughput by 35% compared to previous Node.js services.
- Developed memory-safe, concurrent systems eliminating segfaults and race conditions, reducing production incidents by 40%.
- Led migration of legacy services (Go/Python) to Rust, optimizing CPU usage by 25% and reducing infrastructure costs.
- Implemented distributed caching and rate-limiting mechanisms using Redis and Rust-based workers.
- Collaborated with DevOps to deploy services using Docker + Kubernetes, ensuring 99.99% uptime.
- Mentored junior engineers and conducted code reviews focusing on Rust best practices and performance optimization.
- Engineered OAuth2, TLS, and SQL injection defenses for high-throughput payment backends on Java, Spring Boot, and AWS; safeguarded 2,000+ active users with 99.99% uptime.
- Designed and optimized 15+ production REST APIs (JAX-RS, Hibernate) with PostgreSQL, reducing p99 query latency by 30% under peak load.
- Architected event-driven payment pipelines using Apache Kafka for real-time transaction streaming with guaranteed delivery semantics and fault-tolerant message handling.
- Decomposed monolithic auth and payment services into fault-isolated microservices; expanded JUnit/Mockito test coverage to 85%+, cutting bug resolution time by 25%.
- Automated cloud infrastructure provisioning with Terraform and Docker; optimized Jenkins CI/CD pipeline caching reducing deployment time by 40%.
Projects
Real-time anomaly detection pipeline ingesting three live Kafka streams — server metrics, financial transactions, and IoT sensors. Every event scored by a PyOD ensemble (IForest + LOF + HBOS) with majority voting. Anomalies surfaced via FastAPI REST + WebSocket feed, persisted to PostgreSQL. 44 pytest tests passing.
LLM-powered data analyst agent built on LangChain ReAct. Upload any CSV and ask questions in plain English — the agent picks from three custom tools: PandasTool for queries, PlotlyTool for charts, and StatsTool for outlier detection and trend analysis. Conversation memory enables natural follow-ups. 56 pytest tests passing.
Rust-based AI agent framework implementing the full execution loop (plan → act → observe). Modular tool system with dynamic tool integration, async architecture for multi-step task execution. Published as a reusable library.
Asynchronous data processing pipeline in Rust enabling concurrent execution of multiple stages. Processes 1000+ tasks concurrently with backpressure control and task coordination preventing resource exhaustion.
High-performance concurrent chat server using Tokio handling 1K+ simultaneous TCP connections at sub-100ms latency. Event-driven architecture for real-time broadcasting, stress-tested with simulated clients.
Skills
Education
Certifications
Contact
Let's build
something.
Open to infrastructure, systems, and AI/ML storage roles. Whether it's Rust, distributed systems, or cloud-scale services — let's talk.
Send an Email