Platforms, apps, websites, MT4/MT5 trading systems, and custom AI agent teams - architected, built, and shipped by one developer who also trades live markets with his own automated EAs. Software that doesn't just work. It performs.
Trove Deck Solution is the one-person software company of a full-stack developer and prop-firm trader based in Malaysia. I engineer systems that operate at the intersection of precision and performance. Trading live markets with my own automated EAs taught me that every millisecond matters - and I bring that same discipline to every project I touch.
I don't just build trading tech - I trade with it. My own fully-automated EAs run live on prop-firm capital with news filters, session filters, and risk guards I built myself. MT4, MT5, TradingView, Match Trader - I know the difference between a backtest and a live-safe EA, because my account depends on it.
There is no relay chain to survive. You speak directly with the engineer who architects and builds your solution. Every conversation is technical, every response is precise.
Every engagement begins with a crystalline written scope. Deliverables, boundaries, and change protocols - defined before a single line of code is written. Zero ambiguity at handoff.
I ship battle-tested software - production deployments, comprehensive test coverage, clean architecture. Code that thrives in your infrastructure, not just on a demo screen.
Every inquiry receives a substantive response. I assess feasibility honestly, scope accurately, and deliver consistently. If I can build it, you'll know exactly how. If I can't, you'll know why.
From greenfield web apps to trading automation to AI agent teams that run themselves - one developer scopes it, signs it, and builds it with cinematic precision.
Architected from scratch to your exact specifications. SaaS platforms, internal tools, automation systems, dashboards - when off-the-shelf doesn't cut it, I forge what does.
Production-grade web applications and marketing sites built with Next.js, React, and Node. Clean architecture, fast delivery, real deployment on your infrastructure.
Flutter cross-platform apps running natively on Android and iOS from a single codebase. From concept to app store - or internal distribution for enterprise needs.
API integrations, webhook pipelines, scheduled jobs, scraping, and data workflows. Connect your entire stack and eliminate the manual processes that slow you down.
Production Telegram bots with proper state management, command routing, and webhook handlers. Notification systems, admin tools, and user-facing experiences.
An AI agent embedded in your site that answers visitors instantly and hands off to a human in one tap - routed through Telegram, so several visitors can be helped at once without crossing wires. It learns from every conversation. (The widget on this site runs it.)
Bespoke multi-agent teams engineered to your requirements: a frontier model like Claude Fable or GPT-5.6 as team lead, orchestrating specialist LLMs from different vendors as full-stack developers, reviewers, and testers. They plan, code, cross-review, auto-iterate until acceptance criteria pass, and deliver - with human checkpoints wherever you want them. I run teams like this daily to ship my own products.
Bespoke AI tooling for your team - remote multi-model coding consoles, messaging-to-AI bridges, and multi-LLM routing and orchestration platforms. The same class of tools I build for myself and run every day.
Have a project that doesn't fit a category? Good. Describe what you need - I'll tell you if I can build it. Usually, I can. Versatility is my superpower.
Not a chatbot bolted onto a form - a long-running AI workforce that takes a spec end-to-end: the same pipeline that ships my own products while I sleep, built custom for your company.
The smartest model available - Claude Fable / GPT-5.6 class - does what only top-tier intelligence should: it locks your requirement into a written contract, splits it into parallel work orders, and defines the acceptance criteria every piece will be verified against. Expensive tokens are spent on decisions, not keystrokes.
Work orders fan out to fast, low-cost specialist models from different vendors - full-stack coding, research, data wiring, tests - each on a bounded task with a clear definition of done, running in parallel. Volume work gets produced at a fraction of frontier cost.
Everything returns to the highest-intelligence model wearing its reviewer hat. It diffs the output against the contract, runs deterministic checks - does it build, do the tests pass, does the page actually render - rejects what falls short, and re-dispatches fixes. The loop repeats, unattended, until every criterion is green.
You receive working, deployed software with an item-by-item acceptance report: what passed, what was flagged, what was decided along the way. Human checkpoints wherever you want them - approve-before-deploy, review-only, or fully autonomous.
Most cost optimisation shaves tokens off prompts. The bigger lever is moving the work itself. In this architecture the frontier model touches only the slice of the job where top intelligence changes the outcome - planning and review - while cost-efficient specialists produce the bulk of the keystrokes. In my production pipelines this tiered design has cut spend far deeper than any token-compression trick I've tested, while quality went up - because nothing ships without passing a frontier-level review gate.
I design and build these systems for companies end-to-end: your use-cases, your tools, your compliance constraints. You get the same architecture I run my own business on - a long-running AI workforce that plans, builds, reviews itself, and reports like a senior team.
Not theory - practice. I'm a prop-firm trader, and everything I trade with, I built myself.
It demands engineers who understand platform APIs, execution environments, broker constraints, and the chasm between a working backtest and a live-safe EA. I've crossed that chasm - and I live on the other side of it: as a prop-firm trader, I trade funded capital exclusively with EAs I engineered myself, running high-impact news filters, session and volatility logic, and layered risk guards, fully automated around the clock. Every client build is held to that same standard.
Fully-automated MT4/MT5 strategies built to your specification - entry logic, risk management, execution rules, plus the filters that keep an EA alive in real markets: news blackouts, session windows, spread and volatility guards. The same filter library my own live EAs run on.
Custom technical indicators for MT4/MT5, and TradingView Pine Script indicators & strategy scripts with alerts and backtesting. Precision tools for precision trading - non-repainting where it counts.
Signal delivery systems, broker API bridges, Match Trader custom plugins, and cross-platform integrations between your trading stack and external systems.
Signal provider dashboards, subscriber management, automated delivery via Telegram or REST -full end-to-end signal infrastructure built for scale.
Everything here was built end-to-end by one person - and most of it I run and depend on myself, every single day. A glimpse of what I've shipped.
The EA fleet I trade prop-firm capital with: fully-automated MT5 strategies wrapped in the filter stack that keeps them funded - high-impact news blackouts, session windows, spread and volatility guards, equity protection, and Telegram alerting. Built by me, traded by me, running around the clock.
Trading involves substantial risk. Descriptions reflect how these systems operate, not investment performance or returns. Nothing on this page is financial advice.
A remote console for driving a fleet of AI coding agents from anywhere: multi-vendor LLM routing, live session streaming, background task orchestration, and scheduled automation - delivered as a PWA and native Android app. I ship production code from my phone with it.
Multi-tenant AI customer-service platform: instant AI replies with one-tap human handoff routed through Telegram, Stripe billing, and an embeddable widget that learns from every conversation. The chat bubble on this site is running it right now.
Self-running multi-agent systems: a frontier-model lead (Claude Fable / GPT-5.6 class) plans the work and dispatches specialist LLMs from different vendors as full-stack developers, reviewers, and QA - they cross-verify each other and auto-iterate until acceptance criteria pass. Pipelines like these ship my own products daily; I build them custom for clients.
One developer, deep in every layer. I use the stack that fits the project - no over-engineering, no chasing trends. Just the right tool for the job.
Technical writing for builders -real problems, specific fixes, no filler.
View all posts
Most founders pick the wrong API provider because they optimize for the wrong variable—per-token price instead of total cost per outcome. Here's how to do the math.
Spent three days debugging auth? Here's how Clerk, Auth0, and NextAuth compare—and when to pick each. Plus: when outsourcing auth entirely makes sense for solo founders.
Your frontend is screaming about CORS. You're not broken—you're just missing the one diagram that explains everything. Here's the decision tree that works.
Your app crashes at 3 AM with zero warning. Database fine, code unchanged, but thousands of requests timeout. Connection pool exhaustion is the silent killer most developers discover too late.
Your Next.js blog renders perfectly in production but throws 404 errors on localhost. We break down why dev and production serve static files differently—and show you the one-line fix.
It's 2 AM on Friday and your Next.js build went from 45 seconds to 8 minutes. We walk through the four culprits—dependency bloat, memory exhaustion, cache poisoning, and config drift—plus a 5-minute diagnostic checklist.
Describe what you need. Every message lands directly with the developer - and every inquiry gets a real answer, not a sales pitch.
