Reviewing as Hana Petrova (external)

External reviewer · hana.petrova@example.com

External contactBlind review

DevFlow Conf 2027 — Round 1

AI EngineeringInfrastructure & ScaleProduct & Practice

5 / 18 done

Last active 3 Aug 2026 at 22:30

Everything18 To review13 Done5

From Notebook to Nine-Nines: Productionising ML
To reviewInfrastructure & ScaleProduct & Practice
Evals as a Product Surface: Making Quality Legible
To reviewAI Engineering
Teaching Your Codebase to Answer Questions
To reviewAI EngineeringAgents & Tooling
The Economics of Context Windows
To reviewInfrastructure & ScaleAI Engineering
Fine-Tuning Is a Data Problem, Not a Model Problem
To reviewAI Engineering
Human-in-the-Loop Without Being Human-Hostile
To reviewProduct & Practice
Structured Output Is a Distributed Systems Problem
To reviewAI EngineeringAgents & Tooling
Shipping Agents That Don't Melt Down at 3 A.M.
To reviewAgents & ToolingInfrastructure & Scale
Cutting Inference Costs 70% Without Touching the Model
To reviewInfrastructure & Scale
A Practical Taxonomy of LLM Failure Modes
To reviewAI Engineering
Guardrails at the Edge: Safety Without the Latency Tax
To reviewInfrastructure & Scale
Vector Databases: Choosing One You Won't Regret
To reviewAI EngineeringInfrastructure & Scale
Zero-Downtime Model Rollouts
To reviewInfrastructure & Scale

Evals as a Product Surface: Making Quality Legible

AI EngineeringTalkIntermediate

Evaluations are usually treated as an internal chore. We shipped ours to customers instead, and it changed how the whole team argued about quality.

Speaker details are hidden for this round. Score the proposal on its own merits.

From the application

Have you presented this talk before?: Yes
Where and when did you present it, and what changed since?: Presented an earlier version at a regional meetup last autumn. This version adds nine months of production data and drops the speculative section.
Do you need travel support to attend?: Yes

Your scorecard

Scoring as Hana Petrova (external).

Score 1-5 on relevance, novelty and the speaker's evidence for their claims. Most solid talks land at 3 or 4; reserve 5 for something you would rearrange your day to see. Leave a comment for anything you score 2 or below — we pass anonymised feedback back to submitters.

Score (1 lowest — 5 highest)

Recommendation

Comments for the organizers

Shared with the programme team on the submission page. Never sent to the speaker automatically.

Private note