Data Scientist & Software Engineer · Tashkent, Uzbekistan
I design and build secure, well-architected software — and bring AI automation and multi-agent systems into your processes.
From backend platforms and APIs to assistants that answer from your documents and agents that take on routine work. Clean architecture, security by default, and code your team can run without me.
4+ years building production software · 2+ years in applied AI · Works in English, Russian and Uzbek · Available for projects and consulting
Selected work
Systems that hold up in production.
A few representative projects — architecture, outcomes and what made them robust.
Internal knowledge assistant for a logistics company
A retrieval-augmented assistant that answers staff questions from 4,000+ internal documents with citations — cutting support tickets by a third.
- Python
- FastAPI
- PostgreSQL + pgvector
- Anthropic API
Read case study
Multi-agent workflow for procurement document processing
A supervised multi-agent pipeline that reads supplier offers, extracts structured data, checks it against rules and prepares comparison sheets for human review — cutting per-tender processing from ~2 days to ~3 hours, with every decision left to a person.
- Python
- LangGraph
- FastAPI
- PostgreSQL
Read case study
Backend platform for a vehicle-fleet management product
Designed and built the API platform behind a fleet-tracking SaaS: multi-tenant, event-driven, with clean domain boundaries and 99.9% uptime.
- Python
- FastAPI
- PostgreSQL
- Redis
Read case study
What I build
Web platforms, internal tools, APIs and AI-powered products.
For companies, startups and public-sector teams — from a first version to a system your team runs for years.
AI automation & agents
LLM agents and multi-agent workflows that take routine work off your team's plate — with guardrails, an audit trail and human approval where it matters.
Typical deliverables
- Workflow design
- Agent orchestration
- System integrations
- Accuracy evaluation set
Assistants that answer from your documents (RAG)
Assistants that answer from your own documents and data in Uzbek, Russian or English — accurate, with sources, and safe to put in front of customers or staff.
Typical deliverables
- Ingestion pipeline
- Search & ranking over your documents
- Chat interface
- Quality monitoring
Backend platforms & APIs
The core system behind your product or internal tool — APIs, integrations, accounts and permissions — structured so new features don't break old ones. Python/FastAPI, PostgreSQL.
Typical deliverables
- Architecture document
- OpenAPI-documented API
- Tests & CI/CD
- Deployment runbook
Data engineering & analytics
Pipelines, dashboards and models that turn raw data into answers your team can act on.
Typical deliverables
- ETL / ELT pipelines
- Data models
- Reporting dashboards
- ML prototypes
Every engagement ships with documentation, tests and a handover, so your team can run and extend the system without depending on me.
How I work
A calm, transparent process.
Five steps, each ending in a written output that you keep.
Always included
- Architecture document
- Source code with tests
- CI/CD pipeline
- Deployment runbook
- Security checklist
- Knowledge-transfer session
- Full ownership of code, infrastructure and accounts
Understand
Goals, constraints, users and risks, in your own words. We agree on what "done" means before any code is written.
Output:Scope, risks & estimate
Architect
Modular structure, data model, security model and a delivery plan. Decisions are written down so they can be questioned and revisited.
Output:Architecture document
Build in increments
Typed code, tests and reviews from day one. You see working software every week, not a demo at the end.
Output:Working software, weekly
Harden & verify
Security review, edge cases, load testing and monitoring — plus an evaluation set for the AI parts, so quality is measured, not assumed.
Output:Test & security report
Ship & support
Deployment, documentation and a knowledge-transfer session — then either a clean handover to your team or ongoing support on a retainer. Your choice.
Output:Runbook & handover
Stack
Tools I reach for.
Chosen per project, not by habit — proven and boring where reliability matters, current where it clearly pays off.
- Languages
- PythonTypeScriptSQL
- Backend
- FastAPIPostgreSQLSQLAlchemyRedisCelery / ArqDocker
- AI / LLM
- Anthropic & OpenAI APIsLangGraph / LangChainRAG pipelinespgvector / QdrantEvaluation & guardrails
- Data
- pandasscikit-learnAirflowdbt
- Frontend & Ops
- ReactGitHub ActionsLinuxCaddy / nginx
About
Olimjon, in short.
I'm Olimjon Akhmadjonov, a data scientist and software engineer based in Tashkent. I graduated from INHA University (School of Computer and Information Engineering) and started out as a Python backend engineer, building APIs and services that had to be reliable from day one.
For the last 2+ years I've focused on applied AI — document assistants, agent workflows, AI chat products and data science. What I care about is making it useful in production: answers people can verify, agents that leave decisions with humans, and systems your team can run without me. I apply the same discipline I learned on the backend: clear architecture, security, tests and documentation.
I work directly with businesses, startups and public-sector teams — in English, Russian or Uzbek — as the person who designs, builds and hands over the system. If that sounds like you, I'd be glad to talk.
- Location
- Tashkent, Uzbekistan
- Education
- INHA University in Tashkent · SOCIE
- Languages
- English, Russian, Uzbek
- Availability
- Available for projects and consulting
- Engagements
- Projects · Retainers · Consulting
Experience
From backend to applied AI.
Four roles in four years — from first Telegram bots to production AI systems. The same path the projects above are built on.
July 2024 — present
Tashkent, Uzbekistan
Data Scientist · Datamicron
AI-driven automation: designed and deployed RAG systems, complex LLM workflows and multi-agent architectures (LangChain, LangGraph, LangSmith); prompt engineering for reliable structured outputs; AI data-extraction tools; a chat copilot that builds charts from ontology-driven data sources; shipped as FastAPI and Flask backend services; occasional model fine-tuning.
PythonLangChainLangGraphLangSmithFastAPIFlask
June 2023 — May 2024
Tashkent, Uzbekistan
Python Backend Developer · 4DX
Medical information system for a private clinic with ~10 user roles (doctors, lab technicians, accountants, cashiers, nurses, warehouse staff), each with its own logic and pages; a Telegram bot delivering test results to patients; unit tests for the API; server administration and Docker containerisation.
PythonDockerTelegram Bot API
June — August 2022
Tashkent, Uzbekistan
Backend Developer · Fizmasoft
Two large public-services projects: the public site and a large administrative panel for tracking and managing recruitment processes.
Python
December 2021 — April 2022
Tashkent, Uzbekistan
Backend Developer · ITGO
First role: Telegram bots (e-commerce and HR), administrative panels for websites, landing pages, and CRM/ERP systems for automating business processes.
PythonTelegram Bot API
FAQ
Questions clients ask first.
Short, honest answers. Anything missing — ask the assistant or write to me.
Who owns the code, infrastructure and accounts?
You do — fully. Everything is created in your accounts or transferred at handover: source code, CI/CD, servers, domains and third-party services. Full ownership is a standard deliverable of every engagement, not an extra.
Can you sign an NDA?
Yes. A large part of my work is under NDA — that is why some case studies on this site are anonymised. I can sign yours or provide a standard mutual NDA before we discuss any details.
How long does a project take?
A focused assistant or automation pilot typically reaches production in 8–12 weeks. Larger platforms are delivered in phases, with working software every week — you see progress from week one, not a demo at the end. A precise estimate comes out of the scoping step.
How do we work together remotely?
Remote-first from Tashkent (UTC+5), with comfortable overlap for European and Central-Asian time zones. Weekly demos, decisions in writing, a shared tracker you can open any time. On-site in Tashkent is possible by agreement.
What happens after launch?
Your choice: a clean handover — documentation, runbook and a knowledge-transfer session for your team — or ongoing support on a retainer. Either way the system is built so your team can run it without me.
How is payment structured?
I work as a registered individual entrepreneur in Uzbekistan — with a contract and invoices. An upfront payment starts the work; the rest is split into milestones, each paid on acceptance, so payment follows working results. Currency — whichever suits you. Public-sector procurement procedures are fine.
Can you join an existing codebase or team?
Yes. It starts with a short audit — architecture, security, tests — and a written list of findings. Then we agree what to improve or which features to build. The rule: the system ends up more maintainable than I found it.
In which languages can we communicate?
English, Russian and Uzbek — meetings, correspondence and documentation in whichever your team prefers.
Contact
Let's build something reliable.
Tell me about your project — a few sentences is enough. I reply within 1–2 business days.