I design and ship multi-agent platforms that run in production.
Principal and sole engineer at EM4IT, an AI SaaS product studio (est. 2015). I work with founders and teams in the US and EU who need real infrastructure rather than another prototype.
Most AI products stall before production, and the idea is rarely the reason. Prompt chains, memory layers, retrieval grounding, guardrails, and token economics are architecture problems, and they do not bolt on later.
| Project | What it is | Numbers |
|---|---|---|
| ZENN | Operations platform for beaches and resorts. Guests reserve a specific sunbed from a 3D map of the beach and order food and drinks to it; staff work the same system through a waiter module. Reservations, availability, and payments in one place. | 3D floor plan, Stripe, live |
| Kahani | Therapeutic AI platform with two-tier memory that recalls patients across sessions. Clinical safety guardrails enforced at the orchestration layer. | 17 agents in production |
| Junk Car Boys | Automated vehicle quoting. VIN decode, market ingestion, and a tuned valuation model produce an offer with no human in the loop. | 2,000+ quotes/day, 2.4s avg |
| AI Campus | Adaptive learning platform generating personalized paths from real-time progress analytics. | 5,000+ active students |
| Trading Engine | Multi-agent quant system with BigQuery ML forecasting and automated execution on the Binance API. | 120+ indicators, sub-second |
| Intelligence Engine | Cloud-native NLP pipeline built for throughput and graceful degradation. Enterprise clients including Rio Tinto and Caterpillar. | 1B+ records |
AI: Google ADK, LangChain, multi-agent orchestration, RAG, vector search, Gemini / Claude / OpenAI, evals Full-stack: Python, FastAPI, TypeScript, React, Next.js, Node.js Data: PostgreSQL, BigQuery, Firestore, Qdrant, Redis, MongoDB, Spark Cloud: GCP, AWS, Docker, Kubernetes, Terraform, CI/CD
Building software since before AI meant large language models. CTO and senior engineering roles at companies processing billions of records: Revuze (Senior Data Scientist, distributed NLP at 100M records/day), Binance (Data Specialist), Rio Tinto (Database Specialist), Wagner Equipment / Caterpillar (7+ years, Database Developer), plus 25+ algorithmic trading bots running at 10,000+ RPM.
The through line is throughput. Most of my career has been systems that cannot fall over quietly.
Send requirements, get a working prototype and a fixed-price architecture proposal at no cost. If the fit is wrong, the prototype is yours anyway.
Engagements run 1 to 6 weeks depending on shape: automation systems, MVPs, full AI SaaS platforms, multi-agent systems. Every one ends in a deployed system you own outright, with documentation and a knowledge transfer.
Based in Ulaanbaatar (UTC+8), hours overlapping US Eastern and Pacific. Replies usually within four hours on US business days.
- Studio: em4it.com
- Email: enkhbat@em4it.com
- Upwork: Full work history and client reviews
- Telegram: @golivecash
- LinkedIn: Enkhbat E
