Technical co-founder. On retainer.

I build the thing
you're describing.

Your product is harder than the demo. Let's fix that. I'm Igal — 20+ years building AI and backend systems at real scale. I embed as your technical co-founder and take it to production.

Igal Malisov
20+
Years shipping
0→1
Specialist
10+
AI products
30k
RPS peak traffic

Previously shipped at

BuzzGenie · Ducktor · Aigency.ai · Velocity · Loora · NovoDia · Voyantis · Justt.ai · mesh.security · Zoomin · Wee.bo · Avantis Team
Python Node.js TypeScript Kafka Kubernetes AWS LLM Agents RAG Systems Postgres MongoDB Airflow Spark Docker Redis MLOps GCP Python Node.js TypeScript Kafka Kubernetes AWS LLM Agents RAG Systems Postgres MongoDB Airflow Spark Docker Redis MLOps GCP

Most AI products fail in production —
often for predictable reasons.

I work with a small number of clients at a time. If the left column sounds familiar and the right column sounds like you, we're probably a match.

What goes wrong

  • Prompts & versioning. Every change is a risk; nobody knows what's live vs. tested.
  • Cost & observability. Bills spike overnight with no dashboards, alerts, or ceilings.
  • Reliability. Missing retries and fallbacks — one bad API call takes the pipeline down.
  • Scale. What worked for ten users breaks at a thousand; rebuilds cost more than doing it right once.

This is for you if

  • No technical co-founder. You need someone who owns architecture and outcomes — not ticket intake.
  • Early-stage, real traction. Funding or paying clients; the gap is shipping correctly at the right pace.
  • AI is harder than the demo. LLMs, agents, or data pipelines need to survive real users.
  • Speed without debt. You can't wait six months for a CTO — but the foundation still has to last.

Five ways to work
together.

Depending on where you are and what you need. Every engagement starts with a real conversation — not a sales call.

02 —

Embedded Technical Co-Founder

Ongoing. Fully dedicated to your product. Weekly syncs, async availability, architecture decisions, hiring input, incident response. Senior co-founder leadership — without the equity.

03 —

Full 0→1 Build

I take full technical ownership from whiteboard to production. Stack decisions, architecture, implementation, deployment — everything. I've done this multiple times. The speed comes from already knowing where the traps are.

04 —

AI Product Architecture

For products built on LLMs and AI agents. Prompt versioning, agent orchestration, cost controls, observability, retry logic — the layer between "it works in the demo" and "it works at 2am on a Tuesday."

05 —

Technical Advisory

Monthly retainer. One call per month, async Slack access, architecture and hiring reviews on-demand. For founders who have technical capacity but want a senior voice in the room — someone who's seen the traps and can tell you which ones matter.

What this looks like
in practice.

Four recent engagements — AI infrastructure, platform engineering, fintech architecture, real-time data systems.

Agentic crawl pipeline, vector search & AI chatbot — 0→1

AI Infrastructure · Full Build
RAG Qdrant LangGraph GCP GPT-4o
The problem

Client needed to index and query a large product catalog — no existing pipeline, anti-bot protections on target sites, zero infrastructure to start from.

What I built

Agentic crawler on Cloud Run + Pub/Sub with a 6-layer bot bypass chain. Qdrant vector store with Gemini embeddings. LangGraph RAG pipeline with GPT-4o reasoning and a critic agent for validation. LLM-judge evaluation framework for ongoing quality assurance.

The outcome

40,000 → 1,000,000 items indexed. Crawl time cut from hours to 40 minutes. Full system shipped end-to-end in 30 days.

40k → 1M items indexed. Hours → 40 minutes to crawl. LLM-evaluated quality at every query. Shipped in 30 days.

Production platform rebuilt for 200k DAU

Platform Engineering · Architecture
GCP PostgreSQL Redis Kubernetes Architecture
The problem

12 databases across 3 engines, 7 static UIs running as K8s pods, Redis with no persistence or replication, synchronous DB drivers throughout — scalable on paper, not in practice.

What I did

Consolidated 12 databases across 3 engines down to 5 PostgreSQL databases. Moved static UIs off K8s to GCS + Cloud CDN. Replaced in-cluster Redis with Cloud Memorystore. Migrated sync DB drivers to asyncpg + uvicorn. Decommissioned redundant cluster.

The outcome

Infrastructure cost: $3,400 → $1,500/month. Audio egress reduced 80–90% via CDN. Platform architected to support 200,000 DAU.

$3,400 → $1,500/month. 12 databases across 3 engines → 5 PostgreSQL databases. Platform ready for 200k DAU.

Architecture overhaul for an AI-powered fintech platform

Fintech · Multi-team Architecture
Fintech Architecture Multi-team AI
The problem

Fast-growing AI fintech with 3 independent engineering teams building on diverging stacks — inconsistent patterns, unclear ownership, no unified architecture to scale from.

What I did

Stepped in as Lead Architect across all three teams. Audited the existing system, standardized the stack and service boundaries, and designed a scalable infrastructure model that each team could build against independently.

The outcome

Unified architecture across a multi-team engineering org. Eliminated the divergence risk. Platform positioned to support rapid product growth without architectural debt accumulating across team lines.

3 teams, 1 architecture. Stack standardized, ownership clarified, growth path unblocked.

Real-time data mining engine — 0→1 for a security platform

Security · Systems Engineering
Real-time Data Mining 0→1 Python
The problem

Security platform with no existing data layer — the core product required continuously mining and processing real-time blockchain and on-chain data, at scale, with no prior system to build on.

What I built

Designed and built the fully automated data mining and real-time processing engine from scratch. This became the core technical engine of the entire platform — the subsystem everything else depended on.

The outcome

Core engine shipped and in production. Fully automated, high-throughput, built entirely from zero with no existing infrastructure to inherit.

Core data engine built 0→1. Real-time, automated, in production. Foundation the entire platform runs on.

20 years of building
the real thing.

From gaming platforms to AI-driven fintech. High scale, multiple industries, different kinds of hard.

2024–now
Fractional AI Architect
AI startups
Fractional engagements across AI, backend, and observability. Agentic crawling and RAG pipelines. Production AI architecture and system design. Data pipeline architecture. AI-powered incident response and triage automation. Backend engineering (Python async, AWS, Kubernetes). End-to-end observability stack — deployed to staging and production on AWS.
2024–now
Technical Co-Founder
Ducktor
AI-powered game dev studio. Built Rune Void — a real-time mobile game on Android. → Rune Void on Google Play
2023–2024
Lead Software Architect
Justt.ai
Led system architecture across multiple engineering teams. Designed scalable, maintainable infrastructure for a fast-growing AI-powered fintech platform — standardizing the stack while supporting rapid product growth.
2022–2023
Tech Lead / Architect
mesh.security
Designed and built a fully automated data mining & real-time data processing system from scratch — the core engine of the platform.
2021–2022
Team Leader
Zoomin
Managed 2 teams of 10 across remote and local. Architecture, incident management, stakeholder coordination, and client onboarding.

Earlier: CTO at Wee.bo, Development Manager at Avantis Team — full history on LinkedIn.

Start small.
Commit when it's right.

No long contracts upfront. A scoping sprint means both sides know exactly what we're getting into.

Phase 1

Scoping Sprint

2 weeks · Flat fee

  • 2-hour kickoff — deep-dive into your architecture, codebase, and goals
  • Async Q&A throughout (Slack, same-day responses)
  • Mid-sprint check-in to course-correct
  • Technical assessment of your current state
  • Recommended architecture and tech decisions
  • Roadmap for what needs to be built — and in what order
  • Risk register — what breaks if left unaddressed
Phase 2

Monthly Retainer

Ongoing · 3-month minimum

  • Weekly 1-hour sync — strategy, decisions, unblocking
  • Async availability Sun–Thu, response within 2–3 hours
  • Dedicated Slack channel — direct access, no ticketing
  • Up to 2 critical-issue escalations per month (1-hour response)
  • Monthly written summary — shipped, upcoming, open risks

From people who've
worked with me.

Igal is an exceptional leader. His ability to manage and inspire a team is truly remarkable. He has a unique talent for identifying each team member's strengths and channeling them effectively to achieve outstanding results.

DB
Doron Ben David
CEO & Co-Founder · Indoor Robotics
Verified on LinkedIn ↗

Igal has a rare ability to help teams evolve and grow. He doesn't just solve problems — he helps the people around him become better at solving them too.

MA
Michael Arenzon
Architect · AppsFlyer
linkedin.com/in/arenzon ↗

Think we might be
a fit?

30 minutes. You tell me what you're building and where you're stuck. I'll tell you honestly if I can help — and how.