Led by Egor Dultsev, Principal Engineer. Graftsoft specializes in AI pipelines, fintech scaling, and high-load architecture for companies where failure has a cost.
Small team. Senior execution. No juniors on critical paths.
Specializations
Studio Capabilities
Four service lines, all delivered at principal-engineer level. No hand-offs.
Find the real bottleneck.
We review your current system — codebase, infrastructure, or just a whiteboard design. We'll surface the actual bottlenecks, not the assumed ones, and hand you a written remediation plan.
Demos are easy. Production is not.
LLM integrations that don't fail under load: structured output validation, context management, observability, fallback chains, and async inference. Built for real traffic, not demos.
Built for real money.
Fiat-crypto exchange flows, payment rails, operator dashboards, and compliance-ready transaction lifecycle systems. Real-time WebSocket layers, OTP flows, and fraud-resistant state machines.
The 3× ethos.
We've achieved 3× throughput gains and 30%+ latency reductions through architecture-level work: controlled RPC concurrency, intelligent caching, query optimization, and async job pipelines.
Not sure which service fits? Start with a 30-minute discovery call. No commitment.
Book a discovery callThe Engineer
Egor Dultsev
Principal Engineer & Founder
I don't manage engineers building systems — I am the engineer building the system. 6+ years in production environments where performance, reliability, and correctness aren't negotiable.
My background spans fintech platforms (fiat-crypto exchange flows, payment rails), AI/LLM infrastructure (production inference pipelines, RAG systems, agent workflows), and blockchain integrations (smart contracts, multi-chain routing).
I founded Graftsoft to take on the class of problems that require both architectural thinking and hands-on implementation — often at the same time.
Production stack
B.Sc. Applied Informatics — SPbU
St. Petersburg · 2018–2022
I don't just manage; I architect and ship. Every engagement is hands-on at the principal level — from system design through to production deployment. No account managers, no hand-offs to junior teams.
Years in production
High-load Node.js, distributed systems, fintech infrastructure.
Max performance gain
Achieved on a multi-chain routing layer through architecture-level optimization.
Latency reduction
Backend throughput improvement on a creator SaaS via caching and async processing.
Test coverage shipped
Built automated testing infrastructure on a blockchain platform from scratch.
Career timeline
Lead Full Stack Engineer
AI / Fintech SaaS (NDA)
Full Stack Engineer
Creator SaaS platform (NDA)
Backend / Blockchain Engineer
Multi-chain transaction platform (NDA)
Freelance Full Stack Engineer
Fintech · SaaS · Analytics · 10+ projects
How we work
Every engagement follows the same five-stage model. You always know what stage we're in and what comes next.
30 min · No charge
We talk about the problem — not the solution. What's breaking, what's the cost of failure, and what does "done" look like. I ask hard questions. No sales pitch.
1–3 days · Async
We review what exists: architecture diagrams, codebase, specs, or a napkin design. We profile, we instrument, and we identify the actual bottleneck — not the assumed one.
Deliverable
Bottleneck report
2–5 days · Written
You receive a structured scope document: architecture decisions, tech stack, phased timeline, and a cost estimate with clear line items. No ambiguity. No surprises.
Deliverable
Scope document
Weekly cycles
Iterative delivery with weekly demos. The repo is yours from day one. You can see every commit. We move fast and stay clean — because both matter on production systems.
Deliverable
Weekly demo + commit log
Docs included
Production deployment with monitoring in place. Full technical documentation. Optional async support for 30 days post-launch. You own everything — zero lock-in.
Deliverable
Full documentation + runbook
We talk about the problem — not the solution. What's breaking, what's the cost of failure, and what does "done" look like. I ask hard questions. No sales pitch.
We review what exists: architecture diagrams, codebase, specs, or a napkin design. We profile, we instrument, and we identify the actual bottleneck — not the assumed one.
You receive a structured scope document: architecture decisions, tech stack, phased timeline, and a cost estimate with clear line items. No ambiguity. No surprises.
Iterative delivery with weekly demos. The repo is yours from day one. You can see every commit. We move fast and stay clean — because both matter on production systems.
Production deployment with monitoring in place. Full technical documentation. Optional async support for 30 days post-launch. You own everything — zero lock-in.
Selected work
All under NDA. Tech stacks and outcomes are real. Full details available on request.
OTP-based exchange platform with a full request lifecycle: creation → operator interaction → completion. Embedded real-time chat, LLM-driven user onboarding, and async SMS processing.
The challenge
Sub-second state synchronization across operators and clients, fault-tolerant message delivery, and a fraud-resistant OTP flow under load.
Stack
Backend and analytics infrastructure for a creator monetization platform. LLM pipelines for content generation and onboarding, analytics engines for creator dashboards, and A/B testing infrastructure.
The challenge
High-volume analytics aggregation without blocking the critical request path, and keeping LLM inference costs predictable at scale.
Stack
Multi-chain transaction routing layer with smart contract integrations and high-throughput backend services. Full architecture redesign to eliminate bottlenecks in the transaction processing pipeline.
The challenge
The existing routing service was the single-point bottleneck under load. Required architecture-level changes while keeping the live system stable during migration.
Stack
Let's work together
Serious inquiries only. I respond within 24 hours. Tell me the problem — not the solution.