Work

Selected product work, written as concise technical case studies.

Project 01

Healthcare Analytics & AI Platform

Role: Project Lead
Duration: 1+ year

Problem

Healthcare data teams needed analytics over large-scale datasets and relied on manual, SQL-heavy ingestion configuration work that was slow, repetitive, and error-prone.

Solution

Led development of member prioritization, ROI tracking, and interactive dashboards with server-side pagination and multi-level drill-downs, and built an ETL automation framework using FastAPI and Databricks SQL that generates ingestion configurations and MERGE operations across six interdependent relational tables.

Results

Member prioritization and ROI tracking modulesInteractive dashboards with server-side pagination and drill-downsAutomated generation of ingestion configurationsAI-assisted ETL workflows with human review before writes

Stack

FastAPIDatabricks SQLRAG PipelineChromaDBReact
Project 02

Financial Trading Directory & Transaction Management Platform

Role: Project Lead
Duration: 1+ year

Problem

Admins needed controlled ways to manage firm listings and public-site content, while cashback transactions required proof validation and fraud tracking at scale.

Solution

Led feature planning and full-stack development, implementing role-based access control, content management, and transactional cashback workflows with batch proof validation and fraud-tracking capabilities.

Results

Role-gated admin workflows across the platformTransactional cashback with batch proof validationFraud-tracking capabilities on payout flowsDrag-and-drop management of public-site content

Stack

Next.jsTypeScriptPostgreSQLRBACEmail Services
Project 03

Venue Management & Reservation Platform

Role: Founding Full Stack Developer
Duration: 6 months

Problem

Venue teams needed accurate seating control, responsive reservation management, and a smoother customer-facing booking flow across desktop and mobile contexts.

Solution

Developed venue-management features including interactive drag-and-drop floor-plan editors and service availability controls, and built CRM reservation workflows with dynamic status transitions and integrated Stripe checkout for customer bookings.

Results

Interactive drag-and-drop floor-plan editingCRM reservation workflows with dynamic status transitionsIntegrated Stripe checkout for customer bookingsDedicated mobile and desktop reservation experiences

Stack

ReactTypeScriptStripeResponsive UIReal-Time Data
Project 04

Commerce Voice AI

Role: Full Stack & AI Developer
Duration: Personal project

Problem

E-commerce support needed a voice-first agent that could search catalogs, manage carts, and schedule workflows in real time without exposing customer data or falling to prompt injection.

Solution

Engineered a voice agent using ElevenLabs Conversational AI, Next.js, and TypeScript, orchestrated a tool-calling layer integrating 17 backend microservices for catalog search, cart state, inventory validation, and scheduling, and implemented policy-grounded RAG with secure webhook verification, customer ownership assertions, and visible execution traces.

Results

Low-latency voice plus simultaneous text chatTool-calling across 17 backend microservicesPolicy-grounded RAG with prompt-injection defensesSecure webhook signatures and ownership assertions

Stack

ElevenLabs AINext.jsTypeScriptRAGWebhooks
Project 05

Knowledger AI

Role: Full Stack & AI Developer
Duration: Personal project

Problem

Teams needed accurate, context-grounded answers over their own documents and websites, with retrieval quality beyond plain vector search and a chat surface embeddable on third-party sites.

Solution

Built a multi-tenant SaaS featuring concurrent file chunking engines and recursive web crawlers, deployed a hybrid RAG pipeline combining Pinecone vector search, PostgreSQL full-text indexing, Reciprocal Rank Fusion, and custom Jina reranking, and engineered asynchronous workflows with Inngest, Vercel Blob storage, and Pusher WebSockets.

Results

Hybrid retrieval with RRF and custom rerankingConcurrent chunking and recursive web crawlingAsynchronous processing pipelines with InngestEmbeddable lightweight chat widget for third-party sites

Stack

Next.jsPineconePostgreSQLInngestPusher
Project 06

Agent Flow AI

Role: Full Stack & AI Developer
Duration: Personal project

Problem

Autonomous agent workflows needed reliable orchestration, tenant-isolated data access, and safeguards that keep long-running AI executions observable, repeatable, and auditable.

Solution

Designed a multi-agent orchestration engine with Next.js, TypeScript, and the Vercel AI SDK, developed multi-provider LLM integrations with tenant-scoped data isolation, pgvector RAG lookups, and Zod-validated structured outputs, and built human-in-the-loop validation barriers, SSE streaming state, idempotent execution protections, and tamper-proof audit logs.

Results

Specialized worker agents for analysis and researchTenant-scoped isolation with pgvector RAG lookupsHuman-in-the-loop validation with idempotent executionSSE streaming state and immutable audit logs

Stack

Next.jsVercel AI SDKpgvectorSSEZod
Have a project in mind? Let's talk →