Loading devreal.ai…

talk · community record

Julien Le Dem, Datadog, on Reliable AI — Interview with Alexy

Julien Le Dem, principal engineer at Datadog, on reliable AI at AI By the Bay 2025. A back-end engineer, he recently used Claude Code to build a small web app that visualizes Parquet metadata, out of his comfort zone. Fully reliable AI does not exist yet, everything must be triple-checked; if AI is to produce far more code than we can review, validation and verification frameworks must make sure it introduces no flaws or vulnerabilities. Problems whose solutions are easy to verify will be solved quickly; the harder-to-evaluate ones are where the risk lies.

Watch on YouTube ↗Add or suggest a connection ↗

Connections

7 relationships

featuresApache Parquetproject ↗featuresClaude Codeproject ↗part_ofAI by the Bayevent ↗participated_in · incomingAlexy Khrabrovperson ↗participated_in · incomingJulien Le Demperson ↗representedDatadogcompany ↗transcript_of · incomingTranscript: Julien Le Dem, Datadog, on Reliable AI — Interview with Alexytalk ↗

Provenance

Type: talk
Status: Imported / unclaimed
type: interview
title: Julien Le Dem, Datadog, on Reliable AI — Interview with Alexy
origin: conference
source: bythebay
abstract: Julien Le Dem, principal engineer at Datadog, on reliable AI at AI By the Bay 2025. A back-end engineer, he recently used Claude Code to build a small web app that visualizes Parquet metadata, out of his comfort zone. Fully reliable AI does not exist yet, everything must be triple-checked; if AI is to produce far more code than we can review, validation and verification frameworks must make sure it introduces no flaws or vulnerabilities. Problems whose solutions are easy to verify will be solved quickly; the harder-to-evaluate ones are where the risk lies.
youtube_id: E28iqbEWV3s
youtube_url: https://www.youtube.com/watch?v=E28iqbEWV3s
conference_slug: ai-by-the-bay