np✳
Software engineer · AI systems

NikoPermilovsky.

I reengineer critical workflows
with AI.

Previously at Meta, where I shaped my team’s LLM direction and built automation infrastructure with multimillion-dollar projected impact.
↓ See my work
Open to startup engineering roles

Selected work

At Meta, I contributed meaningful code within one week of onboarding and shaped my team’s LLM direction within six months.

MetaCommunity Operations
System architectureHack / Python / APIs

Reengineering data labeling with LLMs

Reengineered a critical human-driven data-labeling workflow in Meta’s Community Operations, designing and building the infrastructure for LLM-based job routing and closure.

More about this

Design & ownership

Proposed the system, defined its architecture, and owned implementation. Built the configuration layer and APIs that let LLM-based routing and closure support different operational workflows.

Outcome

The system enabled approximately $30M in projected 2026 operating-expense savings.

Jobs + configLLM decisionRoute / close
Simplified conceptual flow
$30MProjected 2026 OpEx
savings enabled
MetaLabeling automation
LLM evaluationEvaluation / Prompts / Audits

LLM automation
& evaluation strategy

Led labeling automation from intake to production and designed my team’s evaluations for Meta’s company-wide comparison of internal and third-party LLMs.

More about this

Automation

Owned the full delivery lifecycle—project intake, prompt iteration, evaluation, audits, and deployment—within a company-wide $120M automation initiative.

Evidence

Designed and ran evaluations against product metrics, human baselines, and gold labels where available, producing evidence for the company-wide model comparison.

Outcome

The automation projects I owned represented approximately $6M in projected 2026 operating-expense savings.

$6MProjected 2026 OpEx
savings across owned projects
MetaDeveloper workflows
Agentic developmentAI skills / Structured prompting

Setting a new onboarding standard

Rebuilt data-provider onboarding around an AI automation skill, taking setup from four hours to approximately five minutes. The workflow became the onboarding standard.

More about this

The build

Designed the skill around structured prompting and an eight-step confirmation flow, giving users a repeatable path through data-provider setup.

~5 minFrom a four-hour
setup process

A bit about me

I’m a software engineer who takes ownership of the technical direction and the work required to deliver it. My experience spans AI infrastructure, model evaluation, internal platforms, and full-stack product engineering.

I’m looking for a startup where I can shape the architecture, stay close to the product, and own important systems from the first proposal through production.

Oct 2025–Aug 2026

Meta

Software Engineer

Designed AI automation infrastructure and built an internal agentic platform with asynchronous jobs and chat. Shipped product improvements across React, GraphQL, and Hack.

Built wellbeing controls across two products supporting 400+ projects and 40M+ weekly labels, collaborating with Reality Labs and GenAI Trust & Safety. Taught AI-first engineering in team sessions shared with a 200-person group.

Apr 2024–Oct 2025

Altonworks

Software Developer

Built a consumer billing API that saved approximately 30 hours of manual verification per month, plus email automation that saved another 10 hours. Delivered a FastAPI permissions app and an asynchronous MicroPython monitoring system with a real-time API and alerts.

2021—2023

University of Washington

B.S. Computer Science · Seattle

Graduated at age 20.

Tools I use

Python / TypeScript / React / Next.js / Hack / FastAPI / PostgreSQL / GraphQL / Docker

LLM evaluation · Agent orchestration · Skills · MCP · RAG · Vector databases

Open to startup engineering roles

Get in touch

Tell me about your team, the role, or what you’re building.

Find me on LinkedIn ↗
Your message goes to my private inbox.