Aviv Shpak

▲
▲
x:0000y:0000p:001%s:Introduce

AI / SYSTEMS / AGENTS DATA SECURITY DELIVERY

Aviv Shpak — AI Engineer

AI Engineer

LLM products.
Agent systems.

I build AI applications end to end: data pipelines, model integration, and the interfaces people use.

Build The Agent, Own The System.

LLM applications, bounded agents, and reliable backends. From Python data pipelines to the product people use.

30SyntheticSpectra checks
3Open-sourceProjects
E2EArchitectureTo Delivery

AI Engineer / Products & Agents

I build LLM applications and the systems that make them useful.

AI engineering / From research to working software

I build AI people can use.

Applied experience

An intelligence platform used by dozens of researchers.

I independently designed and built an AI research workspace, owning the Python ingestion pipelines, databases, LLM processing, and application.

The system turned fragmented source material into structured intelligence, combining extraction rules with LLM analysis and a natural-language research interface.

My background in threat intelligence shapes how I build: trace the evidence, define what the model can do, and keep consequential actions under explicit control.

Discuss my experience on LinkedIn

Explore the code / Spectra

Agent reasoning. Explicit boundaries.

An open-source security automation lab with a bounded investigator, independent policy checks, human approval, and durable execution.

2 typed tools
Tenant-bound evidence reads
4-step budget
Structured proposals and safe abstention
30 contracts
Synthetic policy and investigation checks
Read the agent architecture

Local lab · Simulated external providers · Optional LLM adapter

Open to AI engineering roles · AI products, LLM applications, and agents · Based in Israel

01

LLMs & Agents

LLM IntegrationTool CallingStructured OutputsInformation ExtractionAgent Workflows
02

Backend & Data

PythonGoTypeScriptPostgreSQLConvexSQLite
03

Evals & Reliability

Policy EvaluationsHuman ApprovalTenant IsolationFailure RecoveryObservability
04

Delivery & Interfaces

ReactNext.jsCloudflareDockerPlaywrightCI / CD

From The Model. Into The Product.

Useful AI, reliable systems

Bound the tools. Evaluate the behavior. Verify the effect.

Selected Work / 2025 → Now

Under The Hood.

Explore Spectra

Build For The Edge Cases…

Assume something will fail

An AI Product? Let’s Make It Useful.

Building AI? Let’s Talk.

Connect on LinkedIn

Open to AI engineering roles · LLM applications · Agent systems