Skip to résumé

Sheet 01 · Chad Koslovsky

Build for the second run, not just the first

Software Engineer, AI Infra · Tasklet

A successful run tells me the happy path works. I want to know what an operator can see, what can be retried, and how to recover when a run stops halfway through.

Sheet 02

Cover letter

Why Tasklet

Read full cover letterClose cover letter

The part of Tasklet I want to work on is what happens after a model decides to use a tool. Who is allowed to use it? How do I tell whether the last step took effect? What should be retried, and what needs a person to look at it? I’ve built my own agent harness, and those questions are what keep pulling me deeper into the systems underneath it.

At Front Range Internet, I was the engineer on the other end of those questions when production had a bad night. I ran the NOC, built diagnostics for the people operating the service, and checked that systems actually recovered. At Technomancy IT, I’ve built access-controlled tool integrations and a release workflow that ties an image back to Git, checks the deployed service over HTTP, and keeps database restore verification separate from production.

I’d bring that same operator instinct to Tasklet’s harness. A run should leave enough evidence that another worker or a human can pick it up without guessing what happened.

Chad Koslovsky

Sheet 03

Experience

Technomancy IT LLC · Present

Founder / Product Engineer / Systems Lead

Build access-controlled connectors for SaaS and internal workflows using MCP and authenticated service boundaries. Keep tool access and tenant identity separate across shared services used by clients and internal workflows.

Deep dive

The tools span Slack, Google Workspace, Atlassian, and internal services. Working across client-facing and internal workflows meant preserving distinct access boundaries instead of treating every tool as reachable by every user.

Front Range Internet Inc. · 2015–2020

Senior Systems Engineer / Datacenter Architect

Ran NOC operations and built diagnostics for hosting, colocation, and server workloads. Handled on-call escalations and verified recovery, and worked extensively with Docker.

Deep dive

Held on-call responsibility for hosting infrastructure, built tools for NOC and DevOps staff, and made the evidence for audits and recovery legible to other operators.

Office Evolution · 2014–2015

Systems / Software Engineer

Automated router and firewall provisioning with Node.js and built a Linux-based router to separate office tenants. Mapped individual Wi-Fi access to the right office network.

Deep dive

A shared facility is not a shared tenant. The router mapped individual credentials into the proper office VLAN, and provisioning set bandwidth and access policies per office.

PlumChoice / Everon · 2012–2014

Network & Systems Engineer

Diagnosed network and endpoint problems across managed-service customers, including on-call escalations. Built diagnostic tools for recurring support issues.

Deep dive

I worked directly with customers and handled network and endpoint issues across their environments. Reusable diagnostics made recurring support work easier to hand off.

IBM · 2010–2012

Infrastructure and Systems Focal

Operated shared enterprise infrastructure and coordinated incident and change work across teams. Built an internal backup tool and wrote automation for repeatable operations.

Deep dive

The shared infrastructure required change and escalation coordination across application teams. The internal backup tool was used to save on licensing costs; I also wrote scripts for repeated support work.

Sheet 04

Selected work

Independent product

Technomancy IT Platform

Built a release workflow that tags images from Git, applies SQL migrations, and checks container health and public HTTP responses. A separate PostgreSQL restore check validates the backup checksum and restored tables.

Deep dive

The release workflow records deployment and image identity, retains a bounded set of older images, and checks the application from inside the container and over HTTP. The restore workflow targets a disposable database, separate from production. These are capabilities of the tooling, not uptime or recovery-time claims.

Technomancy IT client

Radar research infrastructure

Provided full-stack and DevOps support to the data-science and AI team on a contactless radar research project.

Deep dive

I helped set up and support the environment used by the data-science team while owning full-stack work for the client. The hardware and model work belonged to other specialists; my job was to make their work practical to build and support.

Sheet 05

Capabilities

  • MCP, mTLS
  • Docker, PostgreSQL
  • Monitoring, diagnostics, escalations
  • Linux, networking, AWS