// projects

Work

Systems I've designed and built — from distributed API gateways to multi-agent AI pipelines.

// experience

Where I've Worked

HSV Digital

Backend Engineer

Dec 2025 — Present

Remote

Built the document-intelligence layer end to end (signed R2 uploads, virus scanning, PDF normalization, vector/graph ingestion) feeding a citation-enforced streaming chat that refuses ungrounded questions. Built an LLM evaluation harness scoring production prompts against live records with AI-judge rubrics, and shipped an event-driven notification and AI coaching-drill platform with idempotent queues and dead-letter handling behind server-evaluated feature flags. Built 40+ backend-for-frontend endpoints for leaderboards, competency heatmaps, and readiness scores over a hierarchical multi-tenant model, and led platform-wide migrations (Prisma 6→7 across 211 files, Next.js 15→16, AI SDK v5→v6).

TypeScriptNode.jsNestJSPrismaPostgreSQLRedisPostHog

Wayground (formerly Quizizz)

Software Engineer

Jan 2025 — Oct 2025

Bengaluru, Karnataka

Built an analytics ingestion pipeline moving 50M+ daily events into BigQuery over Kafka and Pub/Sub, holding end-to-end latency under 30s with deduplication and automatic replay of failed batches. Reduced search latency 78% (to under 280ms across 10M+ records) by partitioning OpenSearch indexes and rewriting high-cost queries. Reduced quiz response latency 42% (2.1s → 1.2s) across 1M+ daily quizzes by adding Redis pipelining to Go and Python services and tuning GCP Cloud Run autoscaling.

GoPythonKafkaPub/SubBigQueryOpenSearchGCPRedis

Mood Analytica

Software Development Intern

May 2024 — Jul 2024

Remote

Developed data-driven features, optimized backends, and built APIs to support analytics workflows.

PythonFastAPIReactPostgreSQL

NuclaDB

Vector Search Engine, Built From Scratch

A vector database written from scratch in Go — HNSW indexing, product quantization, a crash-safe write-ahead log, mmap-backed snapshots, and tenant-isolated multi-tenancy — benchmarked head-to-head against a real Qdrant instance, not wrapped around one.

GogRPCOpenTelemetryPrometheusDockerPython
7432 QPS at ef=10, 16x memory reduction via PQ

OpenSkill

AI-Powered Claude Skill Orchestrator

CLI tool that lets you create, share, and execute reusable AI coding-agent skills without copy-pasting markdown files like it's 2005. Works with Claude, Groq, OpenAI, and Ollama, with sub-100ms local command resolution.

GoRedisPostgreSQLNext.jsTypeScriptDocker
sub-100ms command resolution, 20+ workflow templates

VantageEdge

Multi-Tenant API Gateway

Multi-tenant API gateway in Go that routes requests to tenant-registered origins by subdomain. JWT + scoped API-key auth, per-route rate limiting, per-route Redis cache policies, and origin health checks — instrumented end to end with OpenTelemetry.

GoPostgreSQLRedisOpenTelemetryDockerNext.jsTypeScript
<200ms latency at 10K+ req/day

SyncLayer

Real-Time Collaboration Engine

Trello-ish task board where changes show up instantly for everyone. Built with Go, WebSockets, and Redis — because polling is a crime.

GoRedisPostgreSQLNext.jsTypeScript

SentralQ

Multi-Agent AI Debugging

Paste your broken API logs, get back what went wrong and how to fix it. Multiple AI agents argue about the diagnosis so you don't have to.

PythonFastAPINext.jsTypeScriptTailwindDocker
6s avg inference, 30+ reasoning steps

SynthForce

Multi-Agent Startup Simulation

Multi-agent AI system simulating an entire startup workforce collaborating to generate MVP roadmaps. Groq-accelerated LLMs.

Next.jsFastAPIPostgreSQLTypeScriptDockerAWS

Aegis Agent

AI Code Review Automation

AI code review automation platform with microservices architecture for security audits, performance optimization, and code quality.

PythonFastAPINext.jsTypeScriptDocker

Slanine

AI Toolkit Suite

Comprehensive AI toolkit suite with intelligent tools for streamlining workflows and productivity.

Next.jsPostgreSQLTailwindTypeScriptDocker

Random Web

Curated Web Discovery

Curated collection of odd/nostalgic internet corners. Built with Manya Srivastava. Mood-based, time travel, submissions.

Next.jsTailwindTypeScript

SetUp ChatWebApp

Encrypted Real-Time Messaging

Secure real-time messaging with end-to-end encryption and multimedia support.

ReactNext.jsTailwindTypeScript

QuestMe

Community Discussion Platform

Reddit-inspired community platform with nested comments, voting, and real-time updates via Redis.

Next.jsRedisTailwindTypeScriptReact

// deep dives

Architecture Breakdowns

How these systems work under the hood

NuclaDB

Vector Search Engine, Built From Scratch
Distributed SystemsGolangDatabasesPerformance
Live →

Durability Layer

  • fsync'd write-ahead log before every write
  • mmap-backed snapshot persistence
  • restart replays WAL from last snapshot

Indexing & Compression

  • HNSW graph index
  • Optional product quantization
  • Recall/latency tuned via ef

Multi-Tenancy & API

  • Per-tenant graph, WAL, and snapshot isolation
  • Quotas + rate limiting before requests hit the engine
  • gRPC + REST + CLI
Challenges solved:
Every write fsyncs before returning — 350x slower builds than Qdrant, and it's reported, not hiddenCaught a Qdrant benchmark config bug (full_scan_threshold) that would've made the comparison exact-vs-approximateChaos-tested durability guarantees under crash scenarios

OpenSkill

AI-Powered Claude Skill Orchestrator
CLILLMsGolangDeveloper Tools
Live →

Modular LLM Provider System

  • Go interfaces
  • Groq, OpenAI, Anthropic, Ollama support

Local State & Versioning

  • .claude/skills/ directory
  • SKILL.md rollback history

Skill Composition Engine

  • Extends/includes support
  • Template system
Challenges solved:
Skill format standardization across providersAsync AI generation in CLI contextCross-platform compatibility

VantageEdge

Multi-Tenant API Gateway
Distributed SystemsGolangMulti-TenantSecurity
Live →

Subdomain-Based Routing

  • Tenant-registered origins
  • Origin health checks

Auth & Rate Limiting

  • JWT + scoped API keys
  • Per-route rate limiting
  • Token bucket + Redis

Observability

  • OpenTelemetry tracing end to end
  • Per-route Redis cache policies
Challenges solved:
Cascading failure preventionLow-latency overheadDistributed rate limit synchronization

SyncLayer

Real-Time Collaboration Engine
Real-TimeWebSocketsDistributed SystemsGo
Live →

WebSocket Connection Layer

  • Go Fiber stateful connections
  • Connection pooling

Pub/Sub Message Bus

  • Redis Pub/Sub across instances
  • Fan-out pattern

Optimistic UI & Write Batching

  • Batched PostgreSQL writes
  • Client-side optimistic updates
Challenges solved:
Scaling WebSocket connectionsConcurrent edit conflictsDelivery guarantees

SentralQ

Multi-Agent AI Debugging
AI AgentsObservabilityLLMsAutomation
Live →

Log Ingestion & Normalization

  • FastAPI intake
  • Pandas processing
  • Standard JSON schema

Agent Swarm Orchestration

  • Supervisor + specialist agents
  • LangChain backbone

Synthesized Solution Delivery

  • Aggregated findings
  • Reviewer LLM (GPT-4)
  • Streaming output
Challenges solved:
Multi-agent coordinationHallucination controlVaried log format parsing