I ship AI products from zero to production at a fraction of the usual cost.
I'm a product architect for early stage teams. I decide what to build, design the system, define what done means, and ship it to production with AI-assisted development. At AlphaRexX that cut strategy processing cost by 97% and LLM token use by 92%. I take ownership end to end, so your time goes to growth.
Building AlphaRexX from zero to production proved that I can design elite AI pipelines, but it also taught me that I hate fundraising, B2B sales, and administration. I don't want to run a business. I want to scale AI product engines. I am looking to align with an ambitious founder where I can take total Product Architecture & Delivery Ownership for the next 3 to 5 years.
I don't write raw code; instead, I architect the system workflows and leverage AI-assisted tools to ship production-ready platforms at a fraction of standard development costs. Because I am looking for a true partnership, my target compensation is a sustainable base salary combined with a 4-year equity vesting schedule. I want my skin in the game.
I operate globally as an independent contractor. I am fully set up to onboard immediately through Deel or Remote.com, requiring zero visa sponsorship, legal overhead, or cross-border tax compliance on your end
Case studies
Three decisions behind AlphaRexX.
Each one starts with the situation, then the call I made and what changed.
Business Model & Scalability
Why I Pivoted from Automated Trade Execution to Strategy Testing
Situation
AlphaRexX initially aimed to provide automated trading through proprietary strategies connected directly to users' exchange or broker accounts.
Challenge
The model created two major risks: pressure to make unrealistic profit promises and liquidity constraints when hundreds or thousands of users executed the same signals simultaneously, particularly in low-liquidity crypto assets.
Action
I evaluated the business model against long-term scalability and operational risks, then decided to pivot from automated execution to a strategy creation and backtesting platform.
Result
Shifted the product towards helping traders create, evaluate, and test their own strategies without depending on profit guarantees or large-scale simultaneous trade execution.
Key strengths
Business strategy
Risk assessment
Scalability planning
Strategic decision-making
Product Architecture & Cost Optimization
How I Reduced Strategy Processing Costs by 97% Using DSL
Situation
AlphaRexX relied on expensive LLMs to interpret natural-language trading strategies and generate executable code. This created high processing costs, hallucinations, and inconsistent results.
Action
I redesigned the workflow by introducing a lightweight LLM for strategy understanding and extraction, while a deterministic DSL engine handled code generation. I also introduced a semantic verifier to compare generated rules against the original strategy and a clarification workflow to resolve ambiguous user instructions.
Result
Reduced processing costs to approximately 3% of their original level (97% reduction), token consumption from 370K to 30K per run (92% reduction), and API calls from 6 to 2. The additional validation steps improved strategy interpretation and reliability.
Key strengths
Product architecture
AI workflow design
Cost optimization
Validation systems
Engineering coordination
View the technical blueprint
BeforeExpensive LLM generates executable code
370K tokens per run. 6 API calls. Hallucinations and inconsistent results.
AfterLightweight LLM + deterministic DSL engine
30K tokens per run. 2 API calls. About 3% of the original cost.
01Lightweight LLMStrategy understanding and extraction
The natural-language strategy goes in. Entry, exit and risk logic come out.
02Clarification workflowAmbiguities clarified
Unclear instructions go back to the user.
03Deterministic DSL engineRules built
Same input gives the same rules, in place of LLM-written code.
04Semantic verifierRules verified
Generated rules are compared against the original strategy.
Product Differentiation & Market Research
Identifying a Gap in Natural-Language Strategy Testing
Situation
My product research identified a gap in natural-language trading strategy tools. In my assessment, comparable options were limited, while international alternatives were either expensive or lacked support for complex trading concepts.
Action
I evaluated available alternatives based on pricing and capabilities. Instead of focusing only on basic indicators, I prioritized support for advanced concepts such as Fair Value Gaps (FVG), Order Blocks, and complex combinations of trading conditions.
Result
Positioned AlphaRexX around natural-language strategy creation and testing, with a focus on supporting more complex trading logic at a more accessible cost.
Key strengths
Market research
Competitor evaluation
Product differentiation
User pain-point analysis
Product planning
Services
I take ownership end to end, so your time goes to growth.
Product architecture
I decide what to build, design the system, define what done means, and ship it to production with AI-assisted development.
AI workflow design
A deterministic DSL engine replaced expensive LLM code generation.
Cost optimization
Leveraged a lean network of vetted freelancers and AI tools to replace bloated development costs, cutting product delivery expenses by 30%.
Validation systems
Each part is defined, including what done means, so a developer or an AI tool can build it without guessing.
AI adoption
Trained the team on AI tools and workflows, which raised overall delivery throughput by about 2x.
Delivery
Ran daily standups, prioritized tasks, and oversaw pre-deployment testing on staging before anything shipped.
Experience
AlphaRexX
An AI-powered SaaS platform that turns natural-language trading strategies into structured, backtestable execution bots. It is live at alpharexx.io.
AI Product & Operations ArchitectFounder of AlphaRexX, March 2025 to September 2026
Built it from the ground up, owning product strategy, hiring, budgets and delivery. Defined the architecture, system workflows and UI/UX direction, and led execution to live production deployment.
40%Less server memoryLower memory usage on the same workload.Reduced
60%Faster strategy reportsReport generation time cut for every backtested strategy.Reduced
2 sVoice agent response timeAverage, on Google Gemini Multimodal APIs.Led development and testing
1,000+Strategies testedPlatform testing coordinated across the full set.Coordinated
89%Filtered as non-viableStrategies that failed the platform's filtering during testing.Platform filtering
2xDelivery throughputAbout 2x, after training the team on AI tools.Trained the team
How I work
Research first. Failure cases second. Code last.
Every feature goes through the same sequence before anyone writes code. Most of the cost is decided in the first three steps.
1
Research
How the thing can be built and what it needs: data, APIs, infrastructure.
2
Find where it fails
Drawbacks, downsides, leaks and failure cases, listed before they happen in production.
3
Check the cost
Where cost appears, how to manage it, and how to optimize the design before building.
4
Write the plan and the acceptance criteria
Each part is defined, including what done means, so a developer or an AI tool can build it without guessing.
5
Build and test, failure cases first
The builder focuses on the failure cases, tests them and fixes them immediately. After the build I test again and give the failure cases the most weight.
How I work with code. I architect and specify the system, and AI-assisted tools and specialists write the code. I verify the result against the acceptance criteria.
The pipeline
From a video to a backtest.
A trader finds a strategy explained in a video. The platform turns it into exact rules and tests it against past market data. This is how one run moves through the system.
In plain terms
The strategy goes in as words. The system asks if anything is unclear, turns it into precise rules, checks those rules against what the trader actually described, and only then lets the user test it.
Under the hood
Captions are fetched through a residential IP, because YouTube blocks data center traffic. A lightweight LLM (Gemini) extracts the strategy. A deterministic DSL engine generates the rules in place of LLM-written code, and a semantic verifier compares them with the original strategy. Written strategies skip the caption step.
One run from submission to backtest. A lightweight LLM understands the strategy, and the DSL engine builds the rules from it. A semantic verifier checks them before the user sees anything.
Team and delivery
I hired, trained and ran a remote team of up to six.
At AlphaRexX I recruited and managed engineers, interns and freelancers, and made the final call on every feature added or removed.
Hiring
Leveraged a lean network of vetted freelancers and AI tools to replace bloated development costs, cutting product delivery expenses by 30%.
AI adoption
Trained the team on AI tools and workflows, which raised overall delivery throughput by about 2x.
Delivery
Ran daily standups, prioritized tasks, and oversaw pre-deployment testing on staging before anything shipped.
How a change ships
STEP 1DefineThe feature and the research behind it.
STEP 2SpecifyThe plan and the testing criteria go to the builder.
STEP 3Pull requestThe builder opens one when the work is ready.
STEP 4Review and mergeI review and test it with AI-assisted checks, then merge.
Background
A founder who also knows the market.
I understand the trading side of what I build, which is why AlphaRexX is organized around strategies, backtests and risk rules.
Financial markets
4+ years in financial markets, across trading, spot and derivatives, and risk management.
Used results from 1,000+ tested strategies to tighten the platform's rules and filtering.
Founder
Took a product from zero to live production.
Owned strategy, budgets, hiring and delivery.
Changed the business model after weighing risk and scalability.