Principal Architect — AI Agents · Backend · Distributed Systems

Building systems
that think
& scale

9+ years engineering high-throughput backend platforms with Java, Kafka, and AWS. Now building LLM-powered AI agents and RAG applications that solve real business problems at production scale.

Let's Connect → See AI Work
AI AGENT CORE KAFKA EVENT STREAM API GW REST + gRPC VECTOR DB EMBEDDINGS POSTGRES PRIMARY DB AWS CLOUD EC2 · S3 · Lambda REDIS CACHE LLM GPT · CLAUDE PRODUCTION ARCHITECTURE
⚡
9+
Years Engineering
🛡️
99.9%
Uptime Delivered
🚀
1200x
Latency Reduction
🤖
70%
AI Query Resolution
Ayush Srivastava

Ayush Srivastava

Principal Architect · Exly 📍 Gurugram, India
About

Architect by trade, engineer at heart

I'm a Lead Software Engineer and Principal Architect with deep expertise in distributed backend systems. Currently part of the founding team at Exly, a platform empowering creators in the Indian market.

My craft spans system design, cloud infrastructure, event-driven architecture, and increasingly, AI/LLM-powered applications. I've built Kafka-powered notification pipelines, serverless CDNs, and production RAG systems — always with reliability as the north star.

I believe the next wave of great architecture will have AI agents embedded at its core — and I'm building that future right now.

Languages

Java 21PythonSQLBash

Frameworks

Spring Boot 3.xSpring CloudDjango 5.2Hibernate

Infrastructure

Apache KafkaPostgreSQLRedisMongoDB

Cloud & DevOps

AWSDockerJenkinsNginx

AI / LLM Stack

RAGLangChainVector DBsOpenAI / ClaudeEmbeddings

Architecture

MicroservicesEvent-DrivenDDDCQRS
AI Engineering

Building intelligent systems

Combining deep backend expertise with LLM engineering to build AI agents and RAG applications that operate at production scale — not just demos.

I architect AI systems with the same rigor I bring to distributed backends: fault-tolerant pipelines, observability, cost optimization, and graceful degradation. Every AI feature I ship is backed by real infrastructure — Kafka for event ingestion, PostgreSQL for state, Redis for caching, and vector databases for semantic search.

RAG PIPELINE ARCHITECTURE User Query "How do I set up a webinar?" Embed text → vector [0.23, 0.87, ...] Vector DB cosine similarity top-k retrieval Retrieved Chunks docs, FAQs, guides — score > 0.82 LLM Agent prompt + context → grounded answer guardrails · citations · fallback Grounded Response accurate · cited · hallucination-free Kafka Ingest async doc processing PROMETHEUS · GRAFANA · CLOUDWATCH
01
🧠

Exly Support AI Agent

Production RAG application ingesting Exly's entire knowledge base — help docs, FAQs, onboarding guides — into a vector store. Queries are embedded, matched via cosine similarity, and grounded answers are generated with citations. Deployed with guardrails against hallucination and fallback to human agents.

RAGLLMVector DBPythonKafka
⚡ 70% faster query resolution
02
🔗

Agentic Workflow Engine

Multi-step AI agent framework chaining LLM calls with tool-use — database lookups, API calls, Slack notifications, email drafting. Built with retry logic, token budget management, and structured output parsing. Enables non-technical teams to automate complex support and ops workflows.

AI AgentsTool UseLangChainSpring Boot
🔄 Multi-step orchestration
03
📊

AI Infrastructure Layer

Backend powering all AI features — async document ingestion via Kafka, embedding pipeline with batched processing, vector index management, prompt versioning, and cost tracking per LLM call. Redis-backed caching of frequently-hit embeddings for horizontal scaling.

KafkaRedisPostgreSQLEmbeddings
♾️ Production-grade AI infra
Experience

Where I've built things

May 2024 — Present
Principal Architect
Exly — Founding Team
  • Led architecture of 5 mission-critical infrastructure projects with 99.9% uptime
  • Built LLM-based RAG application — 'Exly Support AI Agent' — reducing query resolution by 70%
  • Designed agentic workflow engine with multi-step LLM orchestration and tool-use capabilities
  • Architected Kafka event streaming across 5 microservices — 1200x latency improvement
  • Designed serverless image CDN with CloudFront + Lambda@Edge — 50% faster page loads
July 2022 — April 2024
SDE III
Exly
  • Built high-throughput bulk Email/WhatsApp campaign service — 75% faster execution
  • Led cloud migrations achieving 60% cost savings without impacting availability
  • Implemented Redis caching layers — 40% reduction in API response times
April 2021 — June 2022
SDE II
Exly
  • Designed the slot management system enabling 4 new SKU types and new revenue streams
  • Engineered payment processing modules — 60ms latency reduction, 15% conversion uplift
Oct 2020 — Mar 2021
Senior SDE
Nagarro
  • Migrated e-commerce order system from monolith to microservices on AWS EKS
  • Designed dual-running strategies for zero-downtime migration
Sep 2018 — Oct 2020
Senior SDE
Wishfin
  • Built Lead Management System processing 1M+ leads/month
  • Implemented event-driven workflows with RabbitMQ
Dec 2016 — Sep 2018
Software Engineer
Dawaibank → TCS
  • Backend development with Spring Boot, JPA/Hibernate, and AWS EC2
Impact

Systems that moved the needle

P0-P3 PARTITIONS EMAIL SVC PUSH SVC SMS SVC LATENCY (ms) 10min <500ms

Kafka Event Pipeline

Event-driven notification infrastructure spanning 5 microservices with Kafka streaming.

1200x latency improvement
Kafka · Java · Spring Boot
S3 ORIGIN EDGE EDGE EDGE EDGE Lambda@Edge

Serverless Image CDN

Globally distributed image delivery with CloudFront, Lambda@Edge, and S3 with edge caching.

50% faster page loads
CloudFront · Lambda@Edge · S3
CAMPAIGN 1M+ msgs KAFKA partition EMAIL WHATSAPP DELIVERED

Bulk Campaign Engine

High-throughput Email/WhatsApp campaign orchestration with millions of messages per batch.

75% faster execution
Kafka · Spring Boot · WhatsApp API
BEFORE $$$ → AFTER $ RIGHT-SIZED COMPUTE + NETWORKING

Cloud Migration & Cost Optimization

Strategic migration to optimized AWS infrastructure with right-sized compute and cost governance.

60% cost savings
AWS · EC2 · RDS · Docker
Expertise

Where I go deep

Backend Engineering

Distributed Systems

Cloud & Infrastructure

AI / LLM Engineering

Contact

Let's build something

Open to Principal Architect and Staff Engineer roles — especially at the intersection of distributed systems and AI engineering. Interested in hard problems, event-driven platforms, and teams building at scale. Based in India, open to relocation.

💬

Drop me a message

Fill out a quick form with your name, email, and message — I'll get back to you as soon as possible.

Contact Me→