Neural Ops Copilot
Agentic SRE assistant with guarded tool use.
A retrieval-grounded operations copilot that proposes remediations, drafts runbooks, and executes approved playbooks behind policy gates.
Next.jsPythonOpenAIPostgresTemporal
Open to selective collaborations
Building calm softwareat the edge of intelligence.
I design and ship —with observability, taste, and product judgment baked in from day zero.
About
Hi, I’m Egehan.
I work at the intersection of applied AI, full-stack engineering, and product design—shipping software that teams actually want to live in.
Most recently I've been building agentic workflows, evaluation infrastructure, and customer-facing surfaces where model behavior is observable and tunable. I care deeply about the boring parts: logging, rollback, permissions, and clear operator UX when things go sideways.
Outside of execution work, I write about shipping with uncertainty, design for dense information, and how to keep ML systems honest in production.
North America
Remote-first
Select travel
Systems thinking
I map failure modes early: latency, cost, hallucination, and human workflow—then design architecture that degrades gracefully.
Product instinct
Great AI feels boring in the best way: predictable when it must be, surprising only when invited. I obsess over that balance.
Craft in the details
Typography, motion, and micro-interactions are not decoration—they are trust signals. I ship interfaces that respect attention.
Work
Featured projects
Selected builds where models, infra, and interface design had to move together—no loose seams.
Latent Design Studio
Generative UI for rapid product iteration.
A canvas-first environment where teams sketch flows, attach constraints, and ship interactive previews synced to a design system.
ReactWebGPURustTailwindFramer Motion
2024
SignalForge Eval
Evaluation harness for multimodal models.
Benchmarking suite for vision-language tasks with regression detection, golden sets, and diffable reports for stakeholders.
TypeScriptPyTorchRayS3Datadog
Trajectory
Experience
A condensed timeline of roles where I owned ambiguous problems end-to-end—from prototype to production metrics.
2023 — Present
Staff AI Engineer
Northwind Labs
Leading agent platforms, safety reviews, and inference cost programs across product and infra.
AgentsEvalsKubernetesGo
2020 — 2023
Senior Software Engineer
Lumen Foundry
Built data-intensive services, ML feature stores, and customer-facing analytics with strict SLOs.
PythonSparkReactGCP
2017 — 2020
Product Engineer
Arc Studio
Shipped design systems and growth experiments for a collaborative creative suite used by global teams.
TypeScriptGraphQLDesign Systems
Toolkit
Tech stack
A pragmatic mix of languages, ML tooling, and infra primitives—chosen for leverage, not novelty.
Languages
TypeScriptPythonGoRust
AI / ML
PyTorchOpenAIAnthropicLangGraphEvalsRAG
Web
Next.jsReactTailwindFramer MotionWebGL
Infra
KubernetesTerraformPostgresRedisKafka
Writing
Notes from the field
Long-form thinking on shipping AI products, interface design under pressure, and evaluation as a first-class surface.
6 min read
Shipping with model uncertainty
How to structure releases when outputs are non-deterministic and stakeholders still need guarantees.
Continue reading4 min read
Designing for attention budgets
Minimal interfaces for dense information—patterns that keep operators in flow during incidents.
Continue readingContact
Let’s build something sharp.
Tell me about the problem space, constraints, and what success looks like. I typically reply within two business days.
Prefer a direct line? Reach me at hello@example.com.