PRODUCT & SYSTEMS ENGINEER

Eldiyar

I take a product from idea to working system and make sure it does not fail, lose money, or wake you at three in the morning.

For six years, I have built systems where mistakes are expensive: payments, crypto operations, financial services, trading systems, and real-time products. I work independently or lead a small engineering team, from requirements and architecture through launch and accountability after release.

Discuss a project

6 years in critical systemsfintech · web3 · trading · realtimeAntalya, Turkey · remote (Europe, CIS, Middle East)full-time or contract

RESULTS

Key results

3,000+ RPS and 100M+ requests a day

Designed a microservice payment core in Encore.ts with Clean Architecture and 99.99% end-to-end uptime.

−90% on-chain fee operating costs

Implemented sweep-on-deposit with programmatic TRON energy rental and dynamic EVM gas top-up.

3M+ multi-chain wallets

Built an event-driven real-time transaction tracking pipeline with deposit splitting and nonce tracking; settlement latency is 2–3 seconds.

−70% load on public RPC nodes

Built a DEX-pool WebSocket monitoring engine for 15+ networks with two-tier Redis caching.

3 hours → 3 minutes per application

Designed the deal lifecycle State Machine and RBAC access model.

10,000+ concurrent WebSocket connections

Built multi-tenant media-session isolation with one Node.js process per room.

<250ms latency and −40% AI API spend

Designed an event-driven OpenAI, Deepgram and ElevenLabs integration bus with caching and deduplication.

−50% false trading signals

Built a six-factor Scoring Engine, Regime Classifier and HTF consensus gate.

30–60 minutes → 10–15 seconds per report

Designed analytics across 12+ microservices using Redis Streams and n8n.

Up to 60,000 RPS

Built a fault-tolerant Apache Kafka and NATS messaging bus.

WHAT I DO

Capabilities

Launching products from scratch.

From the conversation of ‘we need this’ to a production version. I shape requirements, select architecture, build, launch, and improve from feedback.

Critical backend: payments, real time, high load.

Systems that process money and customer events and keep working under load, network failures, and partner errors.

Fintech and crypto payments.

Fund acceptance and distribution, wallets, asset custody, KYC/AML, and fee reduction. I have taken a platform through an external security audit.

Putting AI into products.

Production features for analytics, documents, news processing, and voice and text assistants, with quality and cost controls.

Rescuing projects.

I enter systems that fail, slow down, or have outgrown their architecture. I diagnose, fix, and explain the cause in plain language.

Engineering discipline.

Code reviews, mentoring, and development standards. I have completed 100+ deep reviews and taught middle and senior engineers.

CASES

Projects

CASE 01

Trading system with AI analytics

Challenge: Automated crypto trading lost money to inaccurate signals, while market analysis took up to an hour and decisions used stale data.
My role: Principal product engineer, responsible for discovery, architecture, development, launch, and operations.
What I built: A trading decision system with multi-factor market evaluation and strict risk limits, plus AI agents that analyse markets, news, and forecasts in parallel.
What changed: The system capped risk before a trade, and analytics no longer blocked decisions.

−50% false signals10–15s reports instead of 30–60m3% capital maximum daily loss99.99% availability

stack

Node.js (Bun)TypeScriptPythonPostgreSQL (TimescaleDB)RedisCCXTn8nOpenAI API

technical results

Built a six-factor Scoring Engine with PIT-MAD normalization, a five-state Regime Classifier, and an HTF consensus gate; cut false signals by 50%.
Designed Carver volatility-targeting risk management with 15 pre-trade guards and eight exit scenarios, capping daily drawdown at 3%.
Built an event-driven analytics architecture of 12+ Redis Streams and n8n services; reduced reports from 30–60 minutes to 10–15 seconds.

CASE 02

Crypto payment infrastructure

Challenge: The company lost substantial transfer fees, while payment settlement was slow and unreliable as load grew.
My role: Lead engineer for the payment core and custody layer.
What I built: A payment platform that monitors 15+ blockchain networks, automatically sweeps deposits, and optimizes the fee of each operation.
What changed: Fees fell nearly tenfold, settlement took seconds rather than minutes, and the platform sustained 100M requests a day.

−90% fee costs100M requests/day at 99.99%2–3s settlement3M wallets monitored

CASE 03

Currency exchange platform with document control

Challenge: Crypto-to-fiat applications took three hours by hand, while the platform also had to pass an external security audit.
My role: Full-stack engineer for both backend services and operator UI.
What I built: A system where every deal follows a strict workflow, with KYC/AML, role separation, MFA, and eight-language support.
What changed: Operators stopped being the bottleneck; the platform passed the external audit and entered new regions.

3 hours → 3 minutesexternal audit passed8 language zones

CASE 04

Simultaneous speech translation for international events

Challenge: Thousands of participants needed to hear a speaker in their own language without delay or audio interruption.
My role: Lead engineer for the real-time platform.
What I built: A speech-recognition, translation, and voice-delivery system in 50+ languages for thousands of listeners with hot language switching.
What changed: The product supported live events reliably, while AI-service costs fell 40%.

10,000+ listeners<0.25s latency−40% AI-service costs50+ languages

stack

Node.jsReact.jsWebSocketOpenRestyRedisDeepgramElevenLabsOpenAI

technical results

Built multi-tenant media-session isolation with OpenResty, Redis, and PM2 for stable 10,000+ WebSocket connections.
Designed an event-driven AI-provider bus with caching and deduplication; lowered latency below 250ms and API costs by 40%.
Implemented hot switching across 50+ languages without packet loss or a WebSocket disconnect.

CASE 05

Vending-machine fleet management platform

Challenge: Thousands of machines in different cities required payments, tax reporting, remote configuration, and real-time sales visibility.
My role: Backend engineer and architecture-design contributor.
What I built: The platform core that processes all device events and a bidirectional channel for remote machine configuration.
What changed: Tax reporting became automatic, and the platform had capacity for 10× device growth.

99% transactions automatedup to 60,000 events/sec10× growth headroom

CASE 06

Book digitisation pipeline

Challenge: A publisher needed to digitise and prepare thousands of editions for print without months of manual work.
My role: Full-cycle engineer, from data collection to the finished file.
What I built: An automated collection, data-normalisation, and print-file pipeline that resumes from checkpoints after failure.
What changed: The system replaced work that had required external contractors, with no progress loss on failures.

4,000+ editions10× faster0% progress loss−50% storage costs

PRINCIPLES

How I work with AI

AI speeds up development, but it also makes confident, plausible mistakes. I use these controls to keep it useful.

Specification

Written scope, an explicit ‘what we do NOT do’, and acceptance criteria before the first line of code.

Context

AI needs a clear task and the relevant system context.

Agents

Several AI executors work on separate parts of a task in parallel, following a clear hand-off rule.

Quality checks

Every AI feature has a reference set with known correct answers.

Security

AI receives no broader access than the task needs. Money and client data pass deterministic checks.

Observability

I track what AI did, why it did it, and what it cost. Every decision is auditable.

Cost control

Costs are tracked by function: caching, model selection, and fallback scenarios.

Boundaries

AI is for analytics, text, drafts, and hypotheses. Money, access, and critical operations use deterministic code only.

Accountability

No line reaches production without my review. ‘AI wrote it’ is not an explanation.

GOOD FIT

Who I work best with

Founders from idea to first customers.

You need one person to own the product end-to-end.

Fintech and crypto companies.

Payments, wallets, custody, customer checks, and fee reduction.

Teams that hit a ceiling.

The system is slow, fails under load, or is too risky to change. I diagnose the cause and give you a clear plan.

Businesses that want AI implemented for real.

A product feature with measurable quality, predictable cost, and clear boundaries.

Companies where failure is expensive.

Where an outage means lost money, damaged data, or a conversation with a regulator.

not a fit

I am not a fit if you need an executor with no voice in decisions, or if the task is to fill a deadline with features without considering what happens in six months. I will say when I think the plan is wrong.

ENGAGEMENT

How to work with me

Full-time

Remote, overlapping European business hours.

Project contract

Fixed scope, agreed timeline, and acceptance criteria.

Part-time as lead engineer

Several days per week for architecture, reviews, and team direction.

Audit and system rescue

A short engagement to diagnose the problem and deliver a prioritised plan.

The first step is always a short conversation about the task. Afterwards, I say whether I can take it and, if so, with what estimate.

Discuss a project