Projects | Building Data Systems

End-to-End Diamond V360 Scraping & Normalization

Lead Backend Engineer | 2022 – Present

A headless scraping platform on AWS Lambda that ingests and standardizes complex vendor media assets.

Problem & Solution

Reverse-engineered proprietary v360 players to extract hidden JSON/image assets. Built a serverless Puppeteer system to download, crop, and normalize thousands of assets daily into a consistent S3 structure.

TypeScript AWS Lambda Puppeteer S3 SQS

Three-Stage Media Processing Pipeline

System Designer | 2023 – Present

Orchestrated SQS + Lambda pipeline for scraping, downloading, and normalizing diamond media.

Problem & Solution

Decoupled a monolithic process into three retry-able stages (Scrape, Download, Normalize). improved observability and error recovery for high-volume ingestion flows using EventBridge and CloudWatch.

AWS Lambda SQS EventBridge Serverless

GPU Computer Vision Pipeline (Detectron2)

Architect & Implementer | 2023 – Present

AWS Batch system running PyTorch/Detectron2 models on spot GPU instances for mass image segmentation.

Problem & Solution

Serverless GPUs were too limited. Containerized the model and deployed to AWS Batch (g4dn.xlarge). Enabled cost-effective, large-scale visual QA and feature extraction for hundreds of thousands of diamonds.

PyTorch Detectron2 AWS Batch Docker

Diamond Shape Inference & QA Platform

Lead Engineer | 2021 – Present

ML-powered verification system to catch mislabeled diamond shapes in vendor feeds.

Problem & Solution

Vendor data is often inaccurate. Built a feedback loop where CV models predict shape from images, and SQL logic flags discrepancies against vendor tags for manual review.

Python ML / Classification PostgreSQL Snowflake

Diamond Quality Agents with Gemini

Designer & Prompt Engineer | 2024 – Present

LLM agents that evaluate diamond images and metadata for complex quality traits.

Problem & Solution

Automated qualitative checks (color nuances, cut issues) using Gemini. Designed robust prompts and JSON-output workflows to turn unstructured visual/text data into structured QA flags.

Gemini Prompt Engineering Python LLM Agents

Shopify Catalog Sync & Automation

Backend Engineer | 2021 – Present

Sync engine connecting internal inventory DBs to Shopify via GraphQL and Metafields.

Problem & Solution

Built Lambda-based syncs to create/update products, push rich media (v360s), and map custom attributes to Shopify Metafields. Integrated with Shopify Flow for real-time logic.

Node.js Shopify GraphQL AWS Lambda Workato

Vendor ETL Ingestion & Data Quality

Data Engineer | 2020 – 2024

Hardened ETL pipelines for ingesting messy CSV feeds into Snowflake/Postgres.

Problem & Solution

Implemented strict validation, deduplication, and history tables (SCD) to stabilize reporting. Reduced downstream breakages caused by bad vendor data.

Snowflake PostgreSQL SQL ETL

Large-Scale Processing Architecture Design

Solutions Architect | 2024

Blueprint for processing 400k+ certificates using AWS Batch and Serverless.

Problem & Solution

Designed a scalable architecture (RDS -> SQS -> Batch -> S3) to handle massive backlogs and daily increments of GPU workloads. Included IaC (CDK) and cost modeling.

AWS Batch Architecture S3 CloudFormation

Diamond CSV ↔ Shopify ID Enrichment

Backend Engineer | 2025

High-performance Lambda for enriching large CSVs with Shopify IDs via batch DB lookups.

Problem & Solution

Replaced slow row-by-row checks with a memory-efficient stream parser and batched SQL queries. linking vendor files to Shopify products in seconds required for downstream tools.

Python/TypeScript RDS Optimization Stream Processing

SES Email Attachment Ingestion

Backend Engineer | 2024

Automated extraction of zip files from vendor emails into S3 buckets.

Problem & Solution

configured AWS SES to trigger a Lambda that parses MIME messages, extracts .zip contents, and organizes them in S3. Eliminated manual file handling.

AWS SES Lambda S3 Automation

DepthSnap & MimirMap (R&D)

Computer Vision Engineer | 2024

Depth estimation and mapping tools for GoPro footage using Depth Anything V2.

Problem & Solution

Developed Python libraries and CLI tools to estimate distances and generate map coordinates from monocular video. Packaged as pip-installable open-source libraries.

Python Depth Anything V2 MimirMap Open Source

Selected Projects