Tushar Kasaudhan
Software Engineer
Bengaluru, India·...
About
Hey, I'm Tushar, a backend software engineer who cares about latency, idempotency, and the boring details that keep a service honest after the demo.
I spend most of my time in Python and Go — FastAPI, Gin, Kafka, PostgreSQL, Redis. Ex-Razorpay, Kyndryl. Currently at Flexera, building AI-powered compliance platforms and high-scale backend systems.
I don't ship junk. Maintainability isn't optional. And I build best when reliability, observability, and clean failure modes come first.
Developer Snapshot
- Building production systems.
- Backend + AI platforms.
- Reliability-first engineering.
- Obsessed with clean code.
Projects
Experience
Software Engineer·Flexera Software·Bengaluru
Feb 2026 – Present- Designed and developed an AI-powered compliance platform enabling administrators to create production-ready compliance rules through natural language using AWS Bedrock, Databricks Genie, tool-calling, and multi-stage LLM orchestration, reducing rule creation from engineering sprints to under 5 minutes.
- Built a distributed regulatory compliance platform spanning 9 microservices (FastAPI, gRPC, AWS SQS/SNS, Kubernetes) to execute automated compliance assessments across enterprise regulatory frameworks including DORA, FFIEC, APRA, and NYDFS.
- Architected a multi-tenant rule engine with PostgreSQL, ClickHouse, and Databricks SQL supporting organization-scoped rules, rule versioning, suppression workflows, and secure tenant isolation without impacting downstream processing.
- Optimized large-scale compliance analytics by introducing pre-aggregated ClickHouse summary tables and tiered query execution, reducing dashboard API latency from 55s to 11–17s while significantly lowering query costs on multi-million-row datasets.
- Developed event-driven data pipelines using AWS SQS, SNS, Kubernetes CronJobs, and concurrent Python workers, processing daily compliance assessments with batch ingestion of 100K+ findings into ClickHouse.
- Implemented production-grade backend services using FastAPI, gRPC, SQLAlchemy, and ClickHouse, delivering APIs for compliance dashboards, failed asset analytics, regulatory summaries, and software lifecycle insights.
- Integrated the compliance platform with Flexera's Agentic Mesh to enable secure AI-powered conversations through Flexera One while enforcing deterministic SQL validation and tenant isolation before execution.
Engineer·Razorpay·Bengaluru
Aug 2024 – Jan 2026- Webhook Reliability & Retry Automation: Implemented a failure-aware retry controller for an event-driven webhook system, classifying timeouts, 4xx, and 5xx failures to reduce false webhook disablements by ~40%.
- Resilience & Recovery: Built an auto-recovery workflow to probe paused webhook endpoints and safely re-enable delivery, reducing manual merchant intervention by ~50%; integrated a Dead Letter Queue (DLQ) to ensure durable event handling and controlled replay.
- High-Throughput API Optimization: Optimized payment authorization APIs to sustain 10k+ TPS during traffic spikes by tuning HTTP/database connection pools and enforcing strict timeouts; introduced concurrency limits and backpressure controls, cutting P95 latency by ~35%.
- AI-Assisted Configuration Automation: Developed internal routing and merchant configuration services using Python, FastAPI, and LangGraph, enabling natural-language validation of feature flags and backend configurations.
- Self-Service Enablement Platform: Automated safe configuration workflows to reduce feature rollout time from ~2 days to under 10 minutes, significantly lowering manual intervention and improving deployment safety for internal sales and operations teams.
Associate Software Engineer·Kyndryl India·Bengaluru
Jul 2023 – Aug 2024- Built distributed Python automation pipelines processing 2M+ records/day, reducing manual reporting effort by 80% while improving operational reliability.
- Developed and deployed containerized automation services using Docker and Kubernetes, enabling scalable execution across enterprise environments.
- Automated monitoring and alerting workflows using Dynatrace APIs, reducing false alerts by 60% and improving operational visibility.
- Built Elasticsearch-powered operational dashboards supporting 20+ enterprise customers, enabling faster troubleshooting and data-driven operational insights.
- Authored production runbooks and incident response workflows, improving onboarding efficiency and reducing recurring production issues.
Tech Stack
( select tab to filter )
Writing
tusharkasaudhan.substack.comRedis Demystified: The Secret Weapon of Scalable Systems
A practical tour of Redis — why in-memory storage is so fast, and how caching, sessions, pub/sub, and leaderboards show up in real systems and interviews.
The Secret Sauce Behind Uber’s Real-Time Location System
How Uber tracks millions of vehicles in real time — Kafka streams, H3 geo-indexing, Redis caching, and pub/sub — and what to say in a system design interview.
How Netflix Smartly Solved Personalization at Scale
A deep dive into Netflix’s row-based personalization — specialized microservices, ranking, page assembly, and why splitting the problem beats one giant model.
Designing a Real-Time Bidding / Ad Auction System
What happens behind every targeted ad — latency budgets, auction engines, DSP adapters, caching, and the trade-offs interviewers expect you to reason about.
GitHub Activity
@tushar0305500+ commits in the last year
If you've read this far, you might be interested in backend, distributed systems, or AI platform work.
Let's Talk“Reliability is a feature. Idempotency, retries, and clear failure modes before new surface area.”
— Production engineering
© 2026 All rights reserved.
Bengaluru, India · IST