daemonworks.dev
Quant funds · AI agents · built & operated

Quant fund CTO who builds AI agents—
and keeps them running.

Trading platforms for quant funds, and AI agents that do real sales work. Everything here runs in production every day — not demos.

Start a project → See the work
↓ Download the capabilities deck (PDF)
Systems in production now ★ Top Rated on Upwork · 100% Job Success
daemonworks — process monitor
$ systemctl status --fleet
quant-platformFX & gold · tick data · MT5running
market-datahub12+ venues · 35.9M msgs · 0 droppedrunning
ai-sales-agentsbrowser · email · voicerunning
ai-sales-repemail · LinkedIn · phonerunning
4 units · 0 failedlocal LLMs on own GPUs
What I build
Quant funds & trading
AI agents & automation
01

Quant & trading platforms

CTO of a Polymarket / Kalshi trading fund: market data from 12 venues, execution, and the validation harness that decided which models were allowed to trade. The same platform now runs FX and gold research — 85k lines of Python, tick data, 24/7 on my own servers.

02

AI agents that do the sales job

Browser agents that read pages with vision AI and fill contact forms on hundreds of sites a day. Lead discovery from 8+ sources into a CRM. An AI sales rep that works your own inbox. An AI communication coach that gives feedback on tone, pronunciation and facial expression.

03

Backtesting that does not lie

Real ticks, real costs, out of sample. Most strategies I test get rejected — that is the job. Prop trader twice, so I know the trading side, not only the code.

04

LLMs on your own hardware

Local LLMs on our own GPU servers (Qwen 32B, vision models), plus OpenAI / Claude where it makes sense. Browser automation, scraping, LLM pipelines, CRM and sales systems — n8n / Make when that is enough.

Selected work

Systems I built and run.

Each of these runs in production. Click any image to see it full size.

Hardwired — FX gold quant platform
Quant fund · trading platform

Hardwired — FX & gold quant platform

Research and execution, 24/7 on my own servers

12+market data venues
35.9Mmessages, 0 dropped
85klines of Python

I built it first as the stack of a Polymarket / Kalshi trading fund where I was CTO: market data from 12 venues, execution, and the validation harness that decided which models were allowed to trade. The same platform now runs FX and gold research on tick data — S/R wall detection, trend-integrity analysis, MT5 execution — and backtests that do not lie: real ticks, real costs, out of sample.

Pythontick dataQuestDB / PostgresMT5real-tick backtestsstrategy validation
Market data hub: 12+ venues, 35.9M messages, 0 droppedGold: tick-detected S/R wallsTrend integrity analyzerEURUSD fair-value gaps with entriesStrategy validation: real series vs shuffled
GoGodMode — a sales company run by AI agents
AI company

GoGodMode — a sales company run by AI agents

Discovery, qualification, outreach and CRM — it runs on its own

8+lead sources
1M+websites in the pipeline
24/7runs on its own

Agents find companies from 8+ sources (Google Maps, directories, trade shows, Meta ads, news), qualify them, reach out and track every conversation in a CRM with live dashboards. Local LLMs on our own GPU servers (Qwen 32B, vision models), OpenAI / Claude where it makes sense. I built it and I operate it.

Pythonmulti-agent orchestrationlocal LLMs (Qwen 32B)vision modelsMySQLCRM + dashboards
DashboardThe agentsCompanies foundLead sources: Meta ads
AgentsConnect — AI sales rep
Desktop AI agent

AgentsConnect — AI sales rep

A desktop AI agent that does a junior salesperson's day

3channels: email · LinkedIn · phone
500+companies found & fit-scored
Youapprove what goes out

It works inside the client's own Gmail / Yahoo and LinkedIn: finds companies and decision makers, writes first emails and follow-ups to the client's own rules, reads the replies and drafts the answers, and prepares phone calls. Compliant by design — opt-outs, do-not-contact lists, and the client's approval before anything goes out. A cloud brain plus a desktop app that drives the browser.

Python / FastAPITauri desktop appbrowser automationLLMsGmail · Yahoo · LinkedIn
Calls: ready-to-call scriptMessages: first email and its rulesCompanies found, fit-scoredPeople: decision makersSettings: call compliance
AI sales agents — browser, email and voice
AI agents

AI sales agents — browser, email and voice

Agents that do the sales job

1M+websites in the queue
3agents: browser · email · voice
~360 msvoice response time

The browser agent goes through hundreds of websites a day, reads the page with vision AI, finds the contact form and fills it in — popups, cookie banners and all. The email agent answers the replies; the voice agent talks.

browser automationvision AIPythonLLM pipelinesvoice
Browser agentAgent control1M-site queueDiscovered sitesEmail agentVoice agent
Expert Advisors indicators
MetaTrader 4 / 5

Expert Advisors & indicators

Strategies, indicators, copiers and broker bridges

10+years of MQL4 / MQL5
86completed Upwork jobs
100%Job Success · Top Rated

10+ years of expert advisors, indicators and trade copiers. Non-repainting logic, S/R and channels, fair-value gaps, multi-account execution, broker and exchange integrations, webhooks. The track record is on Upwork.

MQL5MQL4MetaTrader 5MetaTrader 4trade copiersbroker APIs
MT5 terminal: XAUUSDMT5 EA libraryBTC: S/R zones and trend channelsXAUUSD: S/R levels
AI signal copier
Trade automation

AI signal copier

Signals from chat channels straight into broker accounts

2platforms: DxTrade + TradeLocker
1 → Naccounts per signal
AIparser for messy signals

Reads trading signals from chat channels, parses even the messy ones with AI, and places them — multi-TP and SL — on many DxTrade and TradeLocker accounts at once. Licensing and Stripe subscriptions built in.

signal parsingLLMsDxTradeTradeLockerStripe
The copierBroker account settingsSignal parser
Track record

Delivered work, on the record.

Top RatedUpwork status
100%Job Success
4.7 ★36 reviews
86completed jobs
2,500+hours worked
→ See the profile, reviews and work history on Upwork
About

Andrei Gireada.

Quant fund CTO who builds AI agents. I was CTO of a Polymarket / Kalshi trading fund, I was a prop trader twice, and I have written MT4/MT5 expert advisors, indicators and trade copiers for more than ten years.

Since 2001 I run Newrosoft, a software company in Bucharest — mobile and console games (Nintendo, Sony licensed developers) before trading and AI.

Bucharest, Romania — EU hours, overlap with the US morning.

Skills
AI agent developmentLLMsPythonBrowser automationWeb scrapingLead generation automationComputer visionOpenAI APIn8nMQL5MetaTrader 5MQL4Algorithmic tradingTrading automationBacktesting
How I work

One senior engineer. Build and run.

Most automation dies after handoff. I stay on the system, so it keeps earning its keep.

01

Scope

A short, direct conversation to find the highest-leverage thing to automate — and whether it's actually worth building.

02

Build

I ship a working system fast, in production-grade code, with observability wired in from day one.

03

Operate

I run it — monitoring, healing, and improving it on a retainer — so it stays reliable as the world around it changes.

Get in touch

Tell me what you're trying to automate.

If it can run without a human, I can probably build it — and keep it running. Send a couple of lines about the problem and I'll reply personally.

→ andrei@daemonworks.dev