Director of Engineering at Dailymotion. I lead the Ad SDK across Web, Native, and CTV.

From Staff Engineer to Director of Engineering at Dailymotion: what I owned in each role and, where it was measured, the result.

1. 01

   2025 — present · Director of Engineering, Ad SDK

   Three squads, three stacks, one roadmap.

   • —Lead 8 engineers in three squads (Web, Native, CTV) across TypeScript, Kotlin Multiplatform and Swift, with one roadmap built with Product and Business. Also interview product manager candidates.
   • —Led the move to AI-assisted development from September 2025: trialled tools such as Cursor, added AGENTS.md instructions to the repositories, then standardised on Claude Code; in 2026, setting up shared skills to keep token usage down.
   • —The Web team's rewrite of the Web Ad SDK (2025–2026) uses clean architecture (explicit ports and adapters, bounded modules, isolated tests), so each change, by an engineer or an LLM agent, needs less context, uses fewer tokens and can be verified by tests.
   • —Results of the rewrite: +10% ad impressions, +10% technical sell-through rate, −5% time to ad impression, and finer error classification.
   • —Privacy by design in all three SDKs since GDPR enforcement in 2018: the IAB TCF consent string is propagated on Web, and read from device storage and passed to the ad server on Native and CTV; consent rate is monitored on every platform.

   FocusPeople leadership, AI-assisted engineering, architecture and privacy by design across three platforms.

2. 02

   2022 — 2025 · Engineering Manager, Ad SDK Web & Native

   Building and growing the Web and Native teams.

   • —Built the Web and Native Ad SDK teams: led 15 hires, backfills included.
   • —Grew engineers through regular 1:1s, mentoring and skill-development plans: four promoted to senior since 2022, three from mid-level and one from junior through mid-level.
   • —Aligned release standards across the Web and Native stacks, and ran feature rollouts jointly with Product.

   FocusHiring, career growth, delivery standards and collaboration with Product.

3. 03

   2020 — 2022 · Staff Engineer & Tech Lead

   Architecture of the video player's ad stack.

   • —Owned the architecture of the video player's ad stack, keeping it coherent and scalable over the long term.
   • —Main technical contact for Product and Business: turned business requirements into technical specifications the team could build.
   • —Made the trade-offs between new ad features and the player's performance budget.
   • —Drove open-source contributions to vast-client-js and vmap-js, JavaScript parsers for the IAB VAST and VMAP video-ad standards.
   • —Interviewed backend (Go) engineering candidates.

   FocusArchitecture, performance budgets, technical partnership with Product and Business, open source.

1. Sep 2025 – present1 yr 1 mo

   Director of Engineering, Ad SDK

   Dailymotion · Paris

   Lead 8 engineers in three squads (Web, Native, CTV), with one roadmap built with Product and Business.

2. Sep 2022 – Sep 20253 yrs

   Engineering Manager, Ad SDK Web & Native

   Dailymotion · Paris

   Managed the Web and Native Ad SDK teams: hiring, career growth, release standards and rollouts with Product.

3. Jan 2020 – Sep 20222 yrs 8 mos

   Staff Engineer & Tech Lead

   Dailymotion · Paris

   Owned the architecture of the video player's ad stack; open-source contributions to vast-client-js and vmap-js.

4. Dec 2017 – Jan 20202 yrs 1 mo

   Senior Software Engineer

   Dailymotion · Paris

   Joined the video player team to work on its ad integration; that team later became the Ad SDK team.

Previous experience

1. Mar 2015 – Dec 20172 yrs 9 mos

   JavaScript / Full-stack Engineer

   Swaven

   Modernised legacy products onto Node.js and AWS.

2. Aug 2012 – Mar 20152 yrs 7 mos

   JavaScript Engineer

   Adobe

   Built widgets for Adobe Campaign and components for the in-house MVC framework.

3. Dec 2011 – Aug 20128 mos

   JavaScript Engineer

   4D (Wakanda)

   Built components for the MVC framework behind Wakanda Studio, and the Studio's debugger on the Firefox Debugger API.

4. Apr 2009 – Dec 20112 yrs 8 mos

   Software Engineer

   Ubisoft (Owlient)

   Built the payment platform used by all Owlient web games; maintained Haricow, Owlient's open-source PHP framework.

5. Nov 2006 – Apr 20092 yrs 5 mos

   Software Engineer

   Rhapso

   Built Rhapso's web-based ERP for the packaging and printing industry.

EducationDUT Génie logiciel · Université Paris 13 · 2004 — 2006

1. Apr 2026 – present6 mos

   AI photo pipeline for rumesh.photos

   AI-powered photo cataloguing, multi-platform publishing and tag normalisation pipeline.

   Problem

   Cataloguing each photo on rumesh.photos, captioning it for each social platform and keeping the tags consistent is repetitive manual work.

   Solution

   Three workflows on a self-hosted n8n instance and my rumesh.photos API (Node.js, MongoDB, Docker, with an admin back office): enrichment, publishing and tag normalisation.

   Per photo, Gemini 2.5 Flash writes a category, tags and a description, then the Instagram and Bluesky captions in one call. Every two weeks, Gemini 3.1 Flash-Lite merges duplicate tags and harmonises their names, to keep the feed's taxonomy consistent for SEO and navigation.

   Result

   2,700+ photos categorised, and posts published to Instagram and Bluesky with no manual step.

   Engineering choices

   • —Data minimisation: only public photos and their tags are sent to the model.
   • —One Gemini call writes both captions, so each photo is analysed once for both platforms, saving tokens and latency.
   • —Model per task: Gemini 3.1 Flash-Lite, a lighter, lower-cost model, for the text-only tag batch; Gemini 2.5 Flash for the per-photo image work.
   • —Tag normalisation runs as one batch: the model returns an old→new mapping, n8n formats it and batch-updates the photos through the rumesh.photos API, the sitemap drops the old tag URLs, and each run sends a Telegram alert.
   • —Workflows run on a schedule and are throttled to stay under the Gemini API rate limits.
   • —Fully automated by design (personal account, public photos, low-impact and reversible output), with a Telegram alert on every categorisation and every post.

   • n8n
   • Docker
   • Gemini 2.5 Flash
   • Gemini 3.1 Flash-Lite
   • Node.js
   • MongoDB
   • Instagram
   • Bluesky
   • Telegram

   • Live feed on rumesh.photos
   • Posts on Instagram
   • Posts on Bluesky

1. i

   Revenue-critical means calm-critical.

   When the code sits on the revenue path, every release has to be safe to ship and safe to undo. That calls for careful rollouts, observability and reversibility, and it's the right bar for any team I lead.

2. ii

   One roadmap. Three stacks.

   Web, Native, and CTV want to drift apart. The director's job is to keep them telling the same story — same APIs in spirit, same delivery cadence, same definition of done — without flattening what makes each platform itself.

3. iii

   Translate, don't relay.

   Engineering managers don't just relay messages between Product and the team. The work is to translate business intent into specs the team can act on — and to translate technical constraints into trade-offs the business can decide on.

4. iv

   Mentorship is a deliverable.

   I track mentorship the way I track features: skill gaps named, owners assigned, progress visible. The next generation of senior engineers and tech leads doesn't emerge by accident.

5. v

   Performance is a feature of the player.

   An ad SDK that slows the player breaks the deal. We measure ourselves on the player's metrics, not just our own, and we say no to features that don't fit the performance budget.

6. vi

   Make change small and verifiable.

   Clear module boundaries, explicit ports and isolated tests keep every change small enough to understand and to verify, whether an engineer or an LLM agent wrote it. That's what lets AI-assisted development speed a team up without lowering the bar.