AI engineering studio

We build AI that actually ships.

Agents, automations, ML systems and MVPs — designed, engineered and deployed by a team that builds AI developer tooling for a living.

  • Agents & tool calling
  • Context engineering
  • From idea to production
reego · agent run live
  1. $reego run revenue-agent
  2. user“Which accounts churned last week, and why?”
  3. thinkplanning · 3 tools available
  4. toolsql.query("churned_accounts", days=7)42 rows · 180ms
  5. toolcrm.notes(account_ids=[…42])ok · 410ms
  6. thinkclustering churn reasons…
  7. toolslack.post(#revenue, report.md)sent
  8. ✓done in 6.2s · 3 tool calls · $0.011
  • OpenAI
  • Anthropic Claude
  • Gemini
  • Llama
  • Mistral
  • MCP
  • LangGraph
  • Claude Agent SDK
  • PyTorch
  • scikit-learn
  • Hugging Face
  • pgvector
  • FastAPI
  • Node.js
  • Next.js
  • BigQuery
  • Redshift
  • dbt
  • Airflow
  • AWS
  • Google Cloud
  • Vercel
  • ~9 yrsbuilding production software & data systems
  • $80K/yrtooling spend saved with a single AI Slack agent
  • 6+AI products & open-source packages shipped
  • 100%of the code and IP is yours at handover

What we build

From a single automation to a full AI product.

We design, build and run the whole thing — or plug into your team for the hard parts.

LLM Apps & RAG

Copilots, chat assistants and search over your own knowledge — grounded, cited and measurable.

  • Retrieval & vector search
  • Prompt & context design
  • Structured outputs
  • Latency & cost tuning

Machine Learning Engineering

Models taken from notebook to production service — trained, served, monitored and retrained.

  • Classical ML & deep learning
  • Fine-tuning open models
  • Model serving & APIs
  • Monitoring & drift detection

Data Science & Analytics

Turn raw data into decisions: analysis, forecasting, experiments and natural-language self-service analytics.

  • Forecasting & segmentation
  • Experiment design & A/B tests
  • Dashboards & KPIs
  • Text-to-SQL self-service

Data Engineering

Reliable pipelines and warehouses that your models, dashboards and agents can trust.

  • ELT with dbt & Airflow
  • BigQuery / Redshift / Postgres
  • Data quality checks
  • Cost optimisation

AI Security & Evaluation

Eval suites, prompt-injection testing, guardrails and supply-chain scanning for AI-powered systems.

  • LLM eval harnesses
  • Prompt-injection & jailbreak tests
  • Guardrails & policies
  • Malware & dependency scanning

Selected work

We ship our own AI tools, too.

Products, open-source packages and client systems — built with the same stack and standards we bring to your project.

Our own products: Shiftera, Remedium, Suite, Progress, RotaMapping and Curato. All products

All work
ToolCalling.ai screenshot
Product

ToolCalling.ai

The function-calling playground for LLMs

A developer playground for LLM tool / function calling. Design tool schemas and test them against OpenAI-compatible, Llama-compatible and Anthropic models with your own API key.

  • Agents
  • Tool calling
  • LLM
FastData screenshot
Open sourcenpm

FastData

Ask your database questions in plain language

Data self-service for everyone. A Claude agent reads your knowledge base, writes read-only SQL, runs it through your own connector (Postgres, MySQL, SQLite or any MCP SQL server) and answers in Markdown. Born from a text-to-SQL agent we built for a data team.

$ npx fastdata
  • Agents
  • Data
  • Text-to-SQL
DevGuard Scanner screenshot
Open sourcepypi

DevGuard Scanner

Scan untrusted repos for malware before you open them

Fake interview tasks, poisoned forks and hijacked packages run code the moment you open the folder. DevGuard finds those triggers — VS Code tasks, install scripts, git and AI-assistant hooks — follows them to the code they run, and gives a verdict. It never executes anything.

$ pipx install devguard-scanner
  • Security
  • Python
  • CLI
ToolAugmented.ai screenshot
Product

ToolAugmented.ai

The field guide to tool-augmented LLMs

Best practices for training and evaluating LLMs on function calling: schemas, system instructions, outputs, training data, multi-turn flows, tool search and adversarial user cases.

  • Tool calling
  • LLM
  • Fine-tuning
AI Slack agents screenshot
Client workfor Aircall

AI Slack agents

AI Slack apps that replaced costly BI tooling

~$80K / year saved

For Aircall's data team we built AI-powered Slack applications and notification flows so teams could query and act on customer data in near real time — one of them cut internal BI tooling spend by roughly $80K a year.

  • Agents
  • Automation
  • Slack
prollm screenshot
Open sourcenpm

prollm

Prompt any OpenAI-compatible model from your terminal

A tiny CLI to prompt OpenAI-compatible models (OpenAI, Mistral, Claude and more) from the terminal, with named profiles for keys, models and endpoints.

$ npm i -g prollm
  • LLM
  • CLI
  • Node.js

Consulting

Consulting for teams building with AI

Not sure where to start, or stuck between a promising demo and a reliable product? We help founders and engineering teams make the right calls — and then help build them.

AI Engineering

Architecture reviews, model & provider selection, build-vs-buy, cost and latency budgets, and production readiness for LLM systems.

Context Engineering

Design what the model actually sees: retrieval, memory, tool schemas, system prompts and context-window budgets that make agents reliable.

Agent Development

Agent architecture, tool design, MCP, evaluation strategy and failure-mode analysis — before you burn months on the wrong loop.

MVP & Product Ideas

Pressure-test AI product ideas, find the smallest thing worth building and get a concrete technical roadmap with estimates.

AI Strategy & Use Cases

Find the high-ROI automation opportunities in your business and prioritise them by value, risk and effort.

Team Enablement

Hands-on workshops on LLM APIs, tool calling, RAG, evals and agentic coding workflows for your engineers.

60–90 min

Advisory call

Bring your hardest question. Leave with a concrete recommendation and next steps.

Book advisory call
Most popular

1–2 weeks

Discovery sprint

Audit, working prototype and a written roadmap with estimates you can act on.

Book discovery sprint

Monthly

Fractional AI lead

Ongoing architecture, code review and hands-on building as part of your team.

Book fractional ai lead

How we work

Evidence first. Then production.

Short feedback loops, working software every week, and no surprises at handover.

  1. 01

    Discover

    A short call and a written brief. We map the problem, the data and what "done" looks like.

  2. 02

    Prototype

    A working slice in days — real data, real model calls — so decisions are based on evidence.

  3. 03

    Build

    Production engineering with evals, monitoring and weekly demos. Your repos, your cloud.

  4. 04

    Ship & hand over

    Deploy, document and train your team. Optional ongoing support or a fractional retainer.

Kind words

Trusted by data and engineering teams.

I wanted to thank you for your excellent work here. You made a real difference, and the team valued your contributions — especially FastData, the Monitor Pro initiative for Hex cost reduction, and the Slack notification apps. We also appreciated your commitment to a smooth handover.
Chenxi Li
Head of Data Science, Aircall
You have had a huge impact at Aircall. Thank you for your contributions — I would be glad to have you on my team again for our AI initiatives.
Sugandha Arora
Head of Data Engineering, Aircall

About Reego

Senior engineers, not a slide deck.

Reego is a small, senior studio founded by Hasan Yılmaz — a METU computer engineer with ~9 years shipping production systems across .NET, Node.js, Python, data platforms and LLM tooling. We've built AI Slack agents, text-to-SQL self-service, data pipelines and developer tools used in the wild. We work like an extension of your team: your repos, your tools, clear milestones and a clean handover.

  • Ship earlyWorking software in days, with evals and metrics so we know it actually helps.
  • Boring where it countsProven infrastructure underneath, so the AI parts are the only interesting ones.
  • You own everythingYour repos, your cloud, your keys. Documentation and handover included.

FAQ

Straight answers.

Anything else? Just ask — we reply within a business day.

How fast can you build an MVP?

Most AI MVPs reach a deployed, demo-ready state in 3–6 weeks. We start with a working prototype in the first week so you can see real results early.

Which models and providers do you work with?

OpenAI, Anthropic Claude, Google Gemini, Mistral and open-weight models such as Llama — hosted or self-hosted. We pick per use case based on quality, latency, cost and data-privacy requirements, and keep you portable between providers.

Who owns the code and the IP?

You do — 100%. We work in your repositories and cloud accounts whenever possible, and every engagement ends with documentation and a handover.

Can you work alongside our existing engineering team?

Yes, that's most of our work. We join your stand-ups, tickets and code reviews, and leave your team able to own and extend what we built.

How do you price engagements?

Fixed-price for well-scoped builds and discovery sprints, monthly retainers for fractional work, and time-and-materials with a cap where scope is still moving. You always get an estimate before we start.

What about data privacy and security?

We sign NDAs and DPAs, use zero-data-retention endpoints or self-hosted models where needed, and never train on your data. Security reviews and eval suites are part of every production AI system we ship.

Start a project

Have an idea, a messy workflow or a stuck prototype?

Tell us what you're building and what "done" looks like. We usually reply within one business day.

Office Kadıköy, Istanbul
I'm interested in

Opens your email app with the message pre-filled.