MPMartin Pelteshki
LinkedIn
Milan, ItalySoftware engineer @ Lumiwealth

Evidence
to execution.

I build reliable AI and trading infrastructure — from open-source broker runtimes to user-facing products.

View selected workGitHub
WebGPU execution trace

Selected systems

Role → constraint → evidence
Open-source infrastructureNow

Lumibot

A backtestable runtime for AI trading agents and algorithmic strategies across real brokers.

Core maintainer
AI product systemsNow

Lumiwealth / BotSpot

The product layer around trading agents: analysis, execution, interfaces, and measured user journeys.

Product engineering
Independent open source2020–2024

Azule

An iOS app-tweak injection CLI distributed through 89 public releases.

Archived

How I work

Operating principles

Reliability lives in transitions: model to tool, strategy to broker, code to deployment, promise to proof.

01

Measure before optimizing.

Establish comparable baselines, preserve behavior, and keep only improvements demonstrated on representative workloads.

02

Own whole path.

Trace contracts across agents, brokers, cloud runtimes, clients, analytics, and production state instead of patching one surface.

03

Make evidence inspectable.

Tests, traces, public source, and exact deployed state matter more than broad claims or polished demos.

Research

Capital markets
Junior Analyst · Bocconi Students Fintech Society · 2026

Tokenization: Rebuilding Capital Markets on Digital Rails

Research on where faster settlement meets market structure, legal ownership, and institutional adoption.

About

Beyond job title

I build trading software @ Lumiwealth across open source, AI agents, broker integrations, cloud execution, web, and mobile.

Much of my current work sits between Lumibot and BotSpot: broker reliability, agent tools and memory, isolated Python analysis, AWS execution, React and Expo clients, plus PostHog experiments across product and checkout journeys.

Earlier, I built Azule, an iOS tooling project used across 10k+ public release downloads. I am also BAI @ Bocconi, where finance and computation keep meeting in useful ways.

LinkedIn GitHub