Building a 20k Events/sec Webhook Service in Rust
How I replaced a bottlenecked Python service with a Rust/Axum webhook processor. The architecture decisions, Tokio concurrency model, and the migration strategy that kept production running.
> Initializing system...
> Loading profile: Abhinav Nair
> Role: Senior Backend Engineer
> Stack: Python · Rust · AWS · GenAI
> Status: Systems operational ✓
Abhinav
Nair
Senior Backend Engineer
·Affinsys.ai · Bengaluru, INBuilding calm, reliable systems behind chaotic workloads.
Building high-throughput distributed systems in Python & Rust, serving 500k+ users across production microservice environments on AWS.
PythonRustAWSGenAIFastAPIDjangoDockerKubernetesPostgreSQL
// about
Who I Am
Senior Backend Engineer with 3 years of experience building high-throughput distributed systems in Python and Rust, serving 500k+ users across production microservice environments on AWS.
Specialized in event-driven architectures, real-time communication systems, and Generative AI integrations using OpenAI Whisper and Realtime APIs. Proven track record of shipping at scale and leading engineering teams end-to-end.
Distributed Systems
Event-driven architectures with RabbitMQ, NATS, and Redis Streams.
High-Throughput APIs
20,000+ concurrent events/sec with Rust/Axum microservices.
Python & Rust
FastAPI, Django, Axum — cross-language interop via pyO3.
Conversational AI
Production LLM-powered IVR with OpenAI Whisper & Realtime API.
// experience
Work Experience
@ Affinsys.ai — Building conversational AI infrastructure
Senior Software Engineer (Backend)
July 2025 – Present
- ▹Lead backend architecture for 5 distributed microservices in Python (FastAPI, Django) and Rust (Axum), deployed on AWS EKS with PostgreSQL and Redis.
- ▹Built a high-performance Webhook Event Service in Rust/Axum capable of processing 20,000+ concurrent events/sec, replacing a bottlenecked Python-based solution.
- ▹Scaled real-time IVR infrastructure using Kamailio SIPS and RTPEngine on AWS EKS; engineered a custom Python autoscaling worker that reduced idle instance cost by 25% and handled 250+ concurrent calls.
- ▹Improved system reliability by building CI/CD pipelines with GitHub Actions, integrating OpenTelemetry observability, and enforcing test coverage — reducing production incidents by 40%.
- ▹Led a cross-functional team of 7–8 engineers across sprint planning, architecture reviews, code reviews, and production deployments — shipping 10+ major features over 3 months.
Software Engineer (Backend)
July 2023 – July 2025
- ▹Engineered a Campaign Manager platform (Django REST Framework, PostgreSQL/OracleDB, Redis, RabbitMQ) from ground up, cutting setup time from 10-20 mins to under 2 minutes and scaling to 500k+ active users.
- ▹Designed a generalized Task Scheduler using Redis event streams to automate time-based operations across 5+ services, eliminating manual cron dependencies and reducing scheduling latency by 300ms.
- ▹Authored cross-language pub/sub libraries for RabbitMQ and NATS using Rust + pyO3, enabling seamless Python/Rust interoperability and adopted across 15+ internal services.
- ▹Built a production LLM-powered IVR and live calling system integrating OpenAI Whisper and Realtime API via FastAPI/SQLAlchemy, achieving end-to-end voice response latency of 150-200ms.
- ▹Authored 70+ unit, integration, and functional tests with full GitHub Actions automation, achieving 95% code coverage across core services.
Software Engineer Intern
October 2022 – July 2023
- ▹Contributed to the Campaign Manager backend (Django REST Framework), building personalization features powered by LLMs that reduced marketing team workload by 65% and boosted user conversion by 30–42%.
- ▹Executed a full NoSQL to SQL migration from MongoDB to PostgreSQL/OracleDB, improving query performance and enabling relational analytics across campaign data.
// education
B.Tech Computer Science & Engineering
Vellore Institute of Technology
PCM with Computer Science
Bal Bharati Public School, Noida
// skills
Tech Stack
Languages
AI / LLM
Backend & Frameworks
Distributed Systems
Databases
Cloud & DevOps
Observability & Testing
Tools
// writing
Technical Writing
Achieving 150ms Voice Latency with OpenAI Realtime API
A deep dive into building production LLM-powered IVR — from WebSocket streaming to Kamailio SIP routing, and every latency optimization that got us under 200ms end-to-end.
Cross-Language Pub/Sub: Rust + Python via pyO3
Writing RabbitMQ and NATS libraries that work natively in both Rust and Python using pyO3 bindings — and why this pattern eliminated an entire category of serialization bugs across 15 services.
Full posts coming soon
// contact
Get In Touch
Open to discussing backend engineering, distributed systems, and new opportunities.