Ashutosh Satapathy
Backend Engineer building
scalable AI & backend systems.
FastAPI • Go • Kafka •Redis • AWS • AI Voice/Chat Agents
About
Engineering systems that scale
I'm a backend engineer specializing in backend systems, distributed architecture, and AI infrastructure. I build production-grade systems that handle real traffic, real failures, and real scale.
My work spans designing event-driven microservices, building real-time voice AI platforms, and architecting multi-tenant SaaS backends. I think in terms of system boundaries, failure modes, and data flow — not just endpoints and CRUD.
I've built real-time AI voice automation systems handling thousands of concurrent calls with distributed queue architecture, worker orchestration, and auto-recovery.
I care about writing code that other engineers can understand, deploy, and maintain. Clean abstractions, proper observability, and production-ready from the first commit.
Services
What I can build for you
End-to-end engineering services for startups and businesses that need production-grade backend systems.
Backend API Development
AI Integrations
SaaS Development
Real-time Systems
Cloud & DevOps
System Design
Payment Integration
Multi-tenant Architecture
Systems I can build
From concept to production. These are the types of systems I specialize in architecting and shipping.
Featured Work
Production-grade systems
Every project is a case study in scalable architecture, clean code, and production readiness.
AI Chat & RAG Platform
Building a production-grade AI chat platform that lets businesses plug in their own data and get context-aware AI responses. Features retrieval-augmented generation with vector search, streaming token-by-token responses via SSE, document ingestion pipeline for PDFs and web content, conversation memory with session persistence, and multi-tenant API key management.
Results
Engineering Highlights
- RAG pipeline with vector search via Pinecone
- Streaming LLM responses with SSE
- Document ingestion — PDF, web, and plain text
- Conversation memory with session persistence
- Multi-tenant API key management and usage metering
- Rate limiting and cost tracking per tenant
Tech Stack
Webhook Delivery Service
Building a production-grade webhook delivery service — the integration layer every SaaS product needs. Reliable event delivery with exponential backoff retries, HMAC signature verification on every payload, delivery logs with status and latency tracking, fan-out to multiple subscribers, event type filtering, rate limiting per endpoint, dead-letter queue with manual replay, and a real-time dashboard for monitoring endpoint health.
Results
Engineering Highlights
- Reliable delivery with exponential backoff retries
- HMAC signature verification on every payload
- Fan-out — one event, multiple subscribers
- Event type filtering per subscriber
- Dead-letter queue with manual replay
- Real-time dashboard for endpoint health monitoring
Tech Stack
Multi-tenant SaaS Backend
Building a complete multi-tenant SaaS backend that handles everything startups need on day one. JWT authentication with refresh token rotation, role-based access control with org/team hierarchy, Stripe integration for subscriptions and usage-based billing, API key management, tenant-isolated data with row-level security, and invite/onboarding flows.
Results
Engineering Highlights
- JWT auth with refresh token rotation
- RBAC with org/team hierarchy and permissions
- Stripe subscriptions and usage-based billing
- Tenant-isolated data with row-level security
- API key generation and usage metering
- Invite flows and team onboarding
Tech Stack
Flight Booking System
Engineered a complete flight booking system using microservices architecture. Implemented JWT authentication, Redis-based seat locking with TTL, payment gateway integration with idempotency keys, and Kafka-driven async booking workflows. Built an admin dashboard for fleet management and real-time booking analytics via WebSocket updates.
Results
Engineering Highlights
- Microservices architecture with service discovery
- Redis-based distributed seat locking with TTL
- Payment integration with idempotency guarantees
- Kafka-driven async booking pipeline
- WebSocket-based real-time booking updates
- Admin dashboard with analytics
Tech Stack
Experience
Where I've worked
Founding Engineer
OSVI AI
Built and scaled a production-grade AI voice automation platform handling 100K+ calls/day with real-time conversational workflows, distributed scheduling, and event-driven post-call pipelines.
- Capable of processing over 100K calls per day with 100+ concurrency, multi-tenant isolation across 50+ database tables and 200+ API endpoints.
- Reduced outbound call drop rates from 30% to 5% with distributed retry and recovery systems.
- Deployed and managed production infrastructure on AWS (Elastic Beanstalk, EC2, RDS, S3, ElastiCache, Route53, IAM, ACM, CloudWatch).
Key Systems Built
Distributed Outbound Call Queue
High-throughput dispatch engine using Redis sorted sets, Lua scripts, and async worker pools with atomic rate limiting, heartbeat monitoring, automatic recovery, and retry-aware scheduling.
Kafka Post-Call Pipeline
Event-driven pipeline for async transcript analysis, structured data extraction, webhook delivery, Google Sheets sync, and dead-letter queue recovery with idempotent replay.
Campaign Scheduling Engine
Multi-level scheduler with priority queues, retry campaigns, callback scheduling, overnight execution, DND filtering, and follow-up windows synced between Redis and PostgreSQL.
Cross-Session AI Memory
Persistent memory using JSONB-backed session models with contact-level persistence, scoped retrieval, fail-open timeouts, and optimized subquery loading to prevent N+1 issues.
Tech Stack
Software Engineer Intern
Iksha Labs · Gurugram, Haryana
Built and optimized the backend for a session replay and user interaction analytics platform powered by rrweb. Engineered high-throughput event capture, processing, and storage pipelines handling 1M+ daily events from browser sessions across thousands of campaigns.
- Migrated a Go monolith to microservices, enabling horizontal scalability and supporting peak traffic of 100K requests with CPU utilization below 50%.
- Scaled event ingestion to over 1M daily rrweb events using Kafka with real-time consistency and reliable async processing.
- Reduced memory usage from 8GB to 10MB by streaming session events to file before S3 upload — over 99% RAM optimization.
- Boosted authorization throughput by 50% via LRU caching in Redis.
- Integrated unit tests into CI/CD boosting coverage to 90%, and reduced staging test time by 70% with automated k6 load testing.
Key Systems Built
Event Ingestion & Processing Pipeline
High-throughput pipeline for capturing and processing rrweb session replay events from browser clients. Built with Kafka consumers for reliable async ingestion, with streaming file uploads to S3 to handle large session payloads without OOM issues.
Dead Letter Queue & Retry System
DLQ with configurable retry logic for failed event processing, recovering over 95% of transient failures. Ensured no session data loss even under high-traffic spikes.
Rate Limiting & Campaign Control
Redis-based sliding window rate limiter controlling over 10K session captures per campaign. Prevented abuse and ensured fair resource allocation across tenants.
Script Integration Tracker
Redis bitmap-based tracking system with PostgreSQL fallback for monitoring rrweb script installation status across client websites with minimal overhead.
Tech Stack
Tech Stack
Tools I work with
Get in Touch
Let's build something great
Have a project in mind? I'm available for freelance work, consulting, and contract work. Let's talk.