DISCUSS A PROJECT
FEATURED PROJECTS & SOFTWARE · AUSTIN, TX

Featured software projects, case studies, and live systems.

See engineering work across AI search, backend systems, workflow automation, and product engineering.

View projects Discuss a project

FEATURED WORK

ScoutLocal ArchitectureHyperlocal hybrid vector searchStripe Docs SearchCited technical documentation retrievalBullMQ Task QueueReliable background job processingMCP SQLite BridgeMCP database access controlsOpen Source CodeInspect public GitHub repositories
AI SEARCH
BACKEND SYSTEMS
WORKFLOW AUTOMATION
DATABASE DESIGN
API INTEGRATIONS
FEATURED WORK

CLIENT ENGAGEMENT · CASE STUDY

ScoutLocal: hyperlocal hybrid search

A marketplace search that combines semantic retrieval with exact database matching, so people find the right local businesses even when their wording does not match the listing.

Problem

Keyword search over marketplace listings cannot read intent. Someone searching for "family Italian dinner" gets nothing unless a listing happens to use those exact words.

Solution

Built a hybrid search architecture: semantic matching with pgvector finds likely results, PostgreSQL text search keeps them tied to real listing facts, and when semantic confidence is low the system falls back to exact matching rather than guessing.

Read the case study

Architecture

  • pgvector HNSW index
  • PostgreSQL full-text search
  • Node.js (Express) API

ENGINEERING PROJECT · LIVE SYSTEM

Stripe Documentation Search

A documentation search that answers only from the source pages it has indexed and shows the passages it used, so a developer gets a cited answer instead of a list of loosely matching pages.

Problem

Keyword search over a large technical documentation library returns loosely matching pages instead of a direct answer grounded in the actual source material.

Solution

A retrieval service that embeds and indexes roughly 5,000 Stripe API pages and developer guides, searches them semantically, and streams citation-backed answers as they are generated.

Launch demo

Architecture

  • In-memory cosine similarity search
  • Gemini embeddings
  • Token-budget context pruning
  • Server-sent event streaming
  • Node.js (Express) service

ENGINEERING PROJECT · LIVE SYSTEM

BullMQ Distributed Task Queue

A background job service that takes long-running work off the web request, retries it with backoff when it fails, and reports each worker's state, so users are not left with timeouts and silent data loss.

Problem

Long-running tasks such as report generation or email dispatch fail when they are tied to an HTTP request, leaving users with timeouts and silent data loss.

Solution

A decoupled task queue microservice with exponential backoff retries, idempotency and failure handling, concurrency limits, and worker state telemetry.

Launch demo

Architecture

  • BullMQ on Redis
  • Exponential backoff retries
  • Idempotency and failure handling
  • SQLite
  • Concurrency limits
  • Live worker telemetry
  • Node.js

OPEN SOURCE PROJECT · LIVE SYSTEM

MCP SQLite Security Bridge

An open-source proxy that sits between an AI agent and a SQLite database, validates every tool call, and stops write mutations at the boundary, so an agent can query production data without being able to change it.

Problem

Connecting an AI agent directly to a database exposes production tables to prompt-injection attacks and unauthorized write mutations.

Solution

An MCP (Model Context Protocol) security proxy that intercepts JSON-RPC requests, validates each tool call against a schema, and traps write mutations at the proxy boundary.

Launch demo

Architecture

  • Model Context Protocol server
  • JSON-RPC 2.0 validation
  • Zod schema validation
  • Proxy-boundary write controls
  • SQLite
PRODUCTION STANDARDS

What clients receive

Every agreed project engagement is structured around maintainable deliverables built for long-term ownership:

✓ Documented Architecture

Clear system diagrams, data flow schemas, and API contracts.

✓ Production Cloud Deployment

Configured CI/CD pipelines, SSL certificates, and environment secret handling.

✓ Full IP & Code Ownership

100% repository transfer. No proprietary dependencies or vendor lock-in.

✓ Automated Unit & E2E Tests

Comprehensive test suites verifying core user flows and API endpoints.

✓ Setup Runbooks & Docs

Step-by-step local development setup and operational maintenance guides.

✓ Post-Launch Support Window

Dedicated bug-fix and onboarding support following project deployment.

EARLIER CLIENT WORK

The client practice behind this work

From 2010 I ran a client web practice: an onsite pricing API integration, an IDX/MLS property portal, custom themes, storefronts and long-running site operations for small businesses, agencies and local government. The projects above are where that experience is applied now, in TypeScript and backend systems.

View the client work archive  
DISCUSS A PROJECT WITH ANISH