Work & projects

Recent platform work and client results, followed by the wider archive. Each entry identifies the context in which the work was done.

Selected work

Professional work · Current

RestoreAI: production platform and AI infrastructure

Read about my RestoreAI work
Result
Reduced dev/staging AWS spend from roughly $700/day to $200/day through GPU right-sizing, workload placement, and unused-resource cleanup.
Infrastructure ownership
I own production infrastructure on Nebius and dev/staging on AWS, including databases, networking, and GPU-backed AI workloads.
Defined deployment workflows
Built isolated development, staging, preview, and production environments so product engineers can ship through defined workflows. Container build-and-deploy pipelines include health checks and rollback paths.
AI workload operations
I operate GPU-backed video search and summarization workloads, testing model configurations and GPU sizing against requirements with NVIDIA NIM/VSS.
Release safety and recovery
Improved release safety and recovery through database restore validation, vulnerability scanning, and operational runbooks.
Technical SOC 2 Type II preparation
I lead technical preparation and evidence automation across AWS, Nebius, GitHub, and Drata. This is preparation work, not completed certification.

Client engagements · Carybin and Letsdap

Startup infrastructure consulting and migrations

Carybin case studyLetsdap case study
Problem
Managed infrastructure costs had outgrown what these early-stage teams needed.
My contribution
I handled migrations, deployment workflows, recovery procedures, and handover. At Letsdap, I moved application hosting while keeping MongoDB managed.
Outcome
Carybin's core hosting went from about $300/month to €8/month. Letsdap's founder reports roughly 90% lower hosting costs and 65% lower total recurring infrastructure spend.

Personal project

KraiNode: browser-based JSON-RPC tooling

Explore KraiNodeOpen the tool
Problem
Investigating an RPC endpoint often starts with a vague report that requests are slow or failing.
My contribution
I built a browser-based JSON-RPC playground for sending requests and inspecting responses and timings.
Outcome
A repeatable way to test endpoint behavior and turn a reported issue into a request another engineer can inspect.

Project archive

Professional work, client engagements, and independent learning.

Cloud / DevOps / Infrastructure / Compliance

Current

Professional work

RestoreAI: Production Platform and AI Infrastructure

Reduced dev/staging AWS spend from roughly $700/day to $200/day through GPU right-sizing, workload placement, and unused-resource cleanup.

Client engagement

Dailzero: Infrastructure Migration and Operational Handover

Migrated Dailzero's application and PostgreSQL/pgvector workloads, verified a staging restore, and documented deployment and recovery procedures.

Client engagement

Letsdap: Lower Hosting Costs, Managed Database

Moved Letsdap application hosting to Dokploy and file storage to Cloudflare R2 while keeping MongoDB managed.

Professional work

SOC 2 as an Infrastructure Engineering Project (Article)

How I led SOC 2 Type II technical readiness across AWS, Nebius, GitHub, Terraform, and Drata, hardening cloud infrastructure, automating evidence collection, and building a custom Drata integration where no native connector existed.

Personal project

ShipCheck Deployment Readiness Checker

A URL inspection tool reporting HTTPS, reachability, response time, security headers, health endpoints, and inspection confidence.

Client engagement

Carybin: Cloud Infrastructure Migration

Migrated Carybin from managed AWS hosting to Coolify on Hetzner, reducing core hosting from about $300/month to €8/month.

Personal project

IaC with Terraform

Reusable Terraform modules for provisioning and managing cloud infrastructure as code, repeatable, reviewable, and version-controlled.

Personal lab

Stateful Services on Kubernetes (Lab)

A personal Mattermost deployment lab covering Kubernetes configuration and persistent storage.

Personal lab

Monitoring on Kubernetes (Lab)

A personal observability lab with Alloy, Grafana, Loki, and Mimir for collecting and inspecting metrics and logs.

Tutorial

Hosting Fizzy on Coolify

A practical, reproducible guide to self-hosting the Fizzy team communication tool on Coolify, from provisioning to a working, email-capable deployment.

Developer Tools / API Reliability

Personal project

KraiNode: JSON-RPC Playground

A browser-based JSON-RPC playground for testing endpoints, inspecting responses, and investigating request behavior.

Personal project

API / RPC Benchmarking Tool

A tool for benchmarking API/RPC endpoints, measuring latency, throughput, and reliability so you can compare providers and catch degraded performance with data, not guesses.

Open-source contribution

EVM Tools (Open Source Contribution)

Open-source contributions to EVM Tools, developer utilities and tooling, including fixes and improvements that make the tools easier to use.

Open-source contribution

DeFiLlama Chainlist (Open Source Contribution)

Open-source contributions to DeFiLlama's Chainlist, maintaining accurate network metadata and RPC endpoint configuration used by many developers.

Personal project

Get Borrow APY for AaveV3 token

A small tool that queries on-chain APIs to retrieve and display live borrow APY rates for Aave V3 tokens.

Data Engineering

Personal project

EVM ETL pipeline using dlt

An ETL pipeline built with the dlt framework that extracts blockchain data, transforms it, and loads it into a warehouse for analytics.

Personal project

Polygon Finance data pipeline (ELT)

An ELT pipeline that extracts and loads Polygon finance data into a warehouse for downstream analytics and reporting.

Personal project

Socrata API pipeline to GCS and BigQuery

An automated pipeline that pulls data from Socrata open-data APIs and loads it into Google Cloud Storage and BigQuery for analytics and reporting.

Personal project

Onchain Portfolio Tracker

An app that aggregates and tracks your total on-chain portfolio across multiple wallets and networks in one view.

Personal project

Polymarket Opportunity Scanner

A fast MVP that turns public Polymarket order book data into a ranked signal dashboard for spotting complete-set pricing dislocations.

Personal project

Hamoye team project (CKD prediction)

A Chronic Kidney Disease prediction model built as a team capstone, data preparation, model training, and evaluation on Google Cloud.

Blockchain / Node Operations

Tutorial

Debugging & Upgrading Production Nodes (Article)

A guide to upgrading production nodes with near-zero downtime, rolling upgrades, health checks, and recovery workflows. Reliability practices that apply well beyond blockchain.

Tutorial

Production Monitoring for a Starknet Node (Article)

Designed and documented a Prometheus + Grafana monitoring stack for Starknet/Pathfinder nodes, including a reusable dashboard and a timeout troubleshooting runbook.

Tutorial

Running a Tempo RPC Node with Monitoring (Article)

A step-by-step guide to standing up a Tempo testnet RPC node on Ubuntu with monitoring wired in from day one.

Personal lab

HyperEVM Node on Kubernetes (Lab)

A personal Kubernetes lab exploring node deployment, configuration, and storage for a stateful blockchain workload.

Open-source contribution

HyperEVM node (Open Source Contribution)

Open-source contributions to the HyperEVM node, infrastructure, documentation, and deployment tooling improvements.