AITss 2026 Summer School

From Prototype to Production

Building Real-World AI Systems

We've all seen cool AI demos, but turning them into real, reliable products is a different challenge.

June 30, 2026, 10:00–11:20 (Kyiv time, UTC+3)Ivan Franko National University of Lviv
session.info()

$ whoami

Md. Fahim Bin Amin - Software & Machine Learning Engineer

$ session.details()

AITss 2026 Summer School

June 30, 2026, 10:00–11:20 (Kyiv time, UTC+3)

Ivan Franko National University of Lviv

$ topics --count

5 topics covered, from prototype to production

▍

scroll

The talk

From Prototype to Production

A walkthrough of what it actually takes to build and ship AI features in production: how modern AI apps are structured (LLMs, RAG, APIs), the common failure modes around latency, cost, and unpredictable outputs, and practical patterns pulled from real-world systems.

What's covered

Five things that break in production

Anatomy of a production AI app

LLMs, RAG pipelines, and the APIs that glue them together.

The gap between demo and product

Why a working prototype is maybe 20% of the real engineering effort.

Latency & cost tradeoffs

Streaming, caching, model selection, and the economics of scale.

Taming unpredictable outputs

Guardrails, evals, structured output, and graceful failure.

Patterns from real systems

Lessons pulled from shipping AI features that survive contact with users.

By the numbers

6th International Summer School “AI Technology 2026” (AITss-2026)

1281 Participants

participants from 60 countries worldwide registered for the school.

707 Students

students represent 142 Ukrainian and 120 international educational institutions.

317 Academic community

employees of higher education institutions are joining the school.

257 IT community

employees of IT companies from Ukraine and abroad registered for the program.

70 Speakers

speakers from 21 countries confirmed their participation in the school.

Resources

Slides

Loading slides...

Resources

Recording

Ask away

Q&A

A companion reference for everything covered in the talk: 23 topic areas, 54 questions, worked examples, and real-world scenarios.

01 · Architecture & the AI Stack02 · RAG & Retrieval03 · Latency & Performance+20 more
Browse the full Q&A guide

About the speaker

Speaker

Md. Fahim Bin Amin

Md. Fahim Bin Amin

Software & Machine Learning Engineer

Fahim is a software and ML engineer building open-source tools, publishing research, and writing about shipping software that lasts. Based in Dhaka, Bangladesh, he has 6+ years of engineering experience across 300+ projects, 12 published papers, and 5+ conference talks.