Skip to content
View BackendArchitectX's full-sized avatar

Block or report BackendArchitectX

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
BackendArchitectX/README.md
Pranay Kadu — Backend and Distributed Systems Engineer

Selected systems

An exact vector-search engine built from the CUDA kernel upward through C++20, Java JNI and a hardened Spring Boot API. The system keeps indexes resident on the GPU, supports FP32 and FP16-storage modes and includes a reproducible benchmark harness.

Measured evidence: 16.7× lower P50 latency than the project’s scalar CPU reference on 500K × 128 vectors · 1,641–1,662 FP32 QPS at batch 32 · 48/48 benchmark cases passed

Architecture · Benchmark methodology · v1.0.0 release

A Java workshop that develops a distributed transactional key-value store in six deliberate stages: hybrid logical clocks, MVCC, distributed routing, transaction records and intents, clock-uncertainty restarts and serializable conflict prevention.

Engineering focus: transaction ownership · snapshot isolation · write skew · uncertainty windows · read restart · intent resolution

Upstream engineering

Project Engineering problem Status
AutoMQ #3493 Prevent the metrics reporter from starting on controller-only nodes Merged
Trino #30973 Coordinate query failure with transaction-commit ownership Review
Fluss #4263 Make cached RPC connection identity endpoint-aware Review
AutoMQ #3579 Add optional authentication to the built-in Prometheus endpoint Review
Fluss #4230 Support custom Paimon lake-table paths in Spark reads Review

Engineering range

Backend — Java 8–21 · Spring Boot · REST · gRPC · concurrency · asynchronous processing
Data and messaging — PostgreSQL · MySQL · MongoDB · Redis · Kafka · event-driven systems
Reliability and performance — idempotency · rate limiting · circuit breakers · caching · SQL tuning · JFR
Platform and operations — Docker · Kubernetes · AWS EKS · Jenkins · CI/CD · CloudWatch · Splunk · Dynatrace
Validation — JUnit · Mockito · deterministic race tests · benchmark design · production telemetry


Email · GitHub

Pinned Loading

  1. Vortex-CUDA Vortex-CUDA Public

    GPU-accelerated exact vector search engine built with CUDA, C++20, Java JNI and Spring Boot, featuring FP16 storage, batched Top-K search, profiling, benchmarking and production-ready REST APIs.

    Cuda

  2. AutoMQ/automq AutoMQ/automq Public

    Diskless Kafka® on S3. 10x Cost-Effective. No Cross-AZ Traffic Cost. Autoscale in seconds. Single-digit ms latency. Multi-AZ Availability.

    Java 10.7k 769