Hi, I'm Bishal Regmi
Backend developer based in Kathmandu, Nepal. I build APIs that handle real traffic, databases that stay fast under load, and systems that don't page you at 3 am.
Technologies I work with
What I do
API & backend architecture
REST and GraphQL APIs, service boundaries, and data models designed to stay maintainable as traffic and team size grow.
Databases at scale
Schema design, indexing, and search infrastructure — PostgreSQL, Typesense, Redis — tuned to stay fast under real production load.
Cloud & infrastructure
Deploying and operating services on AWS with Docker, thinking through failure modes and rollback paths before they matter.
Payments & real-time systems
Idempotent payment integrations, object storage lifecycles, and real-time features built to survive retries and race conditions.
Recent writing
- Tuning PostgreSQL Autovacuum for High-Write Workloads
Prevent PostgreSQL performance crises in high-write applications. Master autovacuum tuning, worker scheduling, and anti-wraparound to keep your database…
postgresqlautovacuumperformancedatabasetuning - Redis Persistence Wars: RDB, AOF, and Hybrid
Master Redis persistence: RDB, AOF, and Hybrid. Learn fork() copy-on-write, AOF rewrite mechanics, fsync policies, and prevent data loss.
redispersistencebackenddata-losssystem-design - Rate Limiting That Actually Works
Stop outages and protect your APIs. Learn the critical differences between token bucket and sliding window algorithms, and how to build a distributed…
rate-limitingsystem-designredisgoapi-gateway - Scaling Real-Time Chat to 10M Concurrent Users
Learn to design a real-time chat system for millions of users. Master WebSocket scaling, pub/sub architecture, message ordering, and presence detection.
system-designreal-timechatwebsocketskafka - Notification Systems: Fan-Out & Rock-Solid Delivery
Master notification system design: choose fan-out, guarantee delivery, throttle channels, and prevent critical alert failures at scale.
system-designnotificationsmicroservicesqueuesscalability - LLM Eval Is Broken: How to Measure RAG Quality
Stop blindly trusting LLM responses. Learn why BLEU/ROUGE fail for RAG and how to build a production-grade evaluation pipeline for faithfulness,…
llm-evaluationragai-engineeringbackendmachine-learning - Chaos Engineering: Building Resilience, Not Just Features
Learn why breaking things on purpose with Chaos Engineering makes your production systems stronger, faster, and more reliable.
chaos-engineeringreliabilitysrekubernetesmicroservices - LLM Observability: Trace Agents, Track Tokens, Detect Lies
Master LLM observability to debug complex AI agents, optimize token costs, and spot hallucinations with OpenTelemetry and NLI models. A 10x guide.
llm-observabilityopentelemetryai-agentstracingcost-attributionhallucination-detectionnli - The Agent Loop: Tool Use, ReAct, and Memory from Scratch
Learn how to build a production-ready AI agent from first principles: master function calling, ReAct loops, and memory design in Python, avoiding common…
ai-agentsllm-opspythonbackendsystem-design - Why Your 16-Core Database Isn't Multithreading Queries
Your beefy database server has 16 cores, but queries feel slow. Discover why PostgreSQL might not be using parallel execution, and how to fix it.
postgresqlparallel-querydatabaseperformancebackend - Idempotency: The Unsung Hero of Resilient Systems
Discover why idempotency is non-negotiable for distributed systems. Learn to prevent duplicate operations, build resilient APIs, and safeguard critical…
system-designapi-designdistributed-systemsresilienceidempotency - Embeddings: The Silent Scalability Trap
Discover how embedding models, while powerful, can become a critical bottleneck in AI systems. Learn about dimensionality, Matryoshka, and fine-tuning.
ai-engineeringembeddingsllm-opsscalabilitymlops - PostgreSQL's Query Planner: Decoding Optimal Execution
Demystify the PostgreSQL query planner's core decisions. Understand how the cost-based optimizer picks query plans and master EXPLAIN ANALYZE for peak…
postgresqlquery-optimizerexplain-analyzedatabase-performanceinternals - Circuit Breaker: Stop Microservice Cascade Failures Cold
Learn why the Circuit Breaker pattern is vital for microservices, preventing cascade failures with clear states and thresholds. Essential reading.
microservicesresiliencecircuit-breakerbackend-patternsdistributed-systems - RAG Accuracy: The Chunking Problem You're Missing
Boost RAG accuracy by mastering document chunking. Learn token, sentence, semantic, and parent-child strategies to fix common pipeline failures.
ragllmchunkingnlpbackend - Prompt Engineering Isn't Dead, It's Grown Up
Why structured outputs, JSON mode, and programmatic prompt optimization are making prompt engineering reliable, testable, and production-ready.
llmai-engineeringbackendprompt-engineeringdevops - Designing a Notification System: Fan-Out and Delivery Guarantees
Patterns and guarantees for building notification systems that scale: fan-out strategies, retries, idempotency, and delivery guarantees.
notificationssystem-design - PostgreSQL Query Planner Internals: How the Cost-Based Optimizer Chooses a Plan
An overview of PostgreSQL's planner: cost model, statistics, join ordering, and why the wrong plan gets chosen sometimes.
postgresqlbackendperformance - Redis Cluster: The 16384 Slots and the Gossip Beneath
A deep dive into Redis Cluster's internals: hash slots, gossip protocol, resharding, MOVED/ASK redirects, and failover mechanics.
redisclusterdistributed-systemshigh-availabilitycaching