Community Manager - French Speaking, Contract

Why Qdrant dominates 2026 vector search workloads. Qdrant's Rust-native architecture offers superior payload control and memory efficiency, using binary quantization to reduce memory usage by 32 times. Engineering teams are increasingly choosing it over Weaviate and Milvus for complex, high-performance multimodal AI retrieval. The Rust advantage and payload control. Qdrant is built in Rust to provide low latency and memory safety. I see engineering teams leave Weaviate because Qdrant provides superior control over vectors and payloads. While Weaviate provides integrated vectorization and a GraphQL API, its runtime uses more memory at scale. Qdrant's binary quantization technology reduces memory consumption by 32 times and increases retrieval speeds by 40 times, and this implementation has very little loss of accuracy. This compression works well with models from OpenAI, Cohere, and Gemini. If your retrieval depends on restrictive payload filters or explicit index management, you should pick Qdrant over Weaviate to avoid the engineering tax of recreating retrieval behavior within your application logic. Qdrant has reached 7 million downloads and 10,000 users. The company raised $37 million to help developers handle unstructured data like text and images. I find that Qdrant's filtering architecture handles complex constraints better than most engines. The managed cloud offering provides one-click deployments, automated version upgrades, backups, and easy management. Breaking away from Milvus and Weaviate. Engineering teams move away from Milvus when they require the specific tuning that Qdrant provides. Milvus handles billion-scale distributed deployments, but it requires more hands-on infrastructure management. Weaviate remains a strong choice for hybrid search, yet it uses more resources at scale. I find that Qdrant's performance for filtered queries is more consistent. Teams evaluating these tools must look at their specific retrieval failures. If your system returns results from the wrong tenant, you need Qdrant. If your system misses exact phrases, you need Weaviate. If your system cannot scale to billions of vectors, you need Milvus. If your system needs the simplest setup, you need Chroma. You know that switching vector databases involves much more than copying embeddings. The vector database market is crowded. Weaviate raised $50 million, Zilliz raised $60 million, and Chroma secured $18 million. Reddit Engineering tested Milvus, Qdrant, and Weaviate at 1 billion vectors. Qdrant scored highest in that test, but Reddit chose Milvus for better replication and debuggability. Deployment flexibility and cost reality. Pinecone costs $500 a month for its Enterprise plan before you add vectors. One practitioner saw a Pinecone bill climb from $50 to $2,847 per month in just three months. I find the managed Pinecone model frustrating because it lacks any self-hosting option, which creates vendor lock-in. Qdrant provides a hybrid cloud model where you manage the data in your own Kubernetes cluster while Qdrant handles the control plane. This satisfies data residency requirements that Pinecone cannot meet. How much engineering time will your team lose during the next migration? Migration involves more than copying vectors. You must map index parameters and validate query behavior. One practitioner reported that migrating between databases takes between 40 and 200 engineering hours, depending on the complexity of the schema. Qdrant allows you to switch deployment options without the risk of proprietary lock-in.

Reddit is ramping up its war on marketing slop. Sep 23, 2026, 5:00 AM PT The company exclusively told CMO Insider that it's taking new steps to combat spam and keep human conversations at the center of its platform. First, it'll prevent suspected spam from appearing in the main home feed while it's under review. Reddit said in testing, this approach cut spam in the home feed by 28%, user reports of spam in the feed by 11%, and push notifications with spam by 7%. It plans to test this approach in search results next. Reddit is also taking a few other steps: * It plans to start explaining to users why it suspects certain content is spam, rather than just showing "Removed by Reddit" * It won't recommend posts or comments from accounts that have been asked to verify they're human or register their automation, but haven't yet done so "What's made Reddit really unique over the last 21 years, and special, is the humanity that you see on Reddit every day," said Adam Collins, Reddit's chief communications officer. "Our goal is to preserve that for the future as well. That's really what these efforts are about." Reddit said it had uncovered a coordinated network of 485 accounts that created more than 44,000 pieces of content across the platform, often about work and productivity. These accounts presented themselves as regular users but were, in fact, connected to businesses, violating Reddit's rules requiring marketing activity to be disclosed. Reddit said it reversed the potential benefit of the coordinated behavior by removing karma and votes that resulted from the manipulation. Karma is a score on Reddit that measures a user's reputation. The spam problem is a particular threat to Reddit, which considers its human-moderated communities to be its greatest advantage over other social platforms. Reddit vs. AI. After going public in 2024, Reddit increased advertising on its forums and entered into partnerships with OpenAI and Google. Marketers took note and have been racing to create content for Reddit. Brands particularly hope their efforts will help them show up in AI search, where Reddit is highly cited. The problem: Self-promotional content from brands, especially if they use AI to make it, risks harming the human quality that Reddit sees as setting itself apart from other platforms. A July report from the AI detection firm Pangram found that Reddit was the least affected by AI among five platforms, with 3% of its short-form content flagged as AI. LinkedIn was at the high end at 30%. Reddit has been proactive. In July, Reddit announced tools to combat AI-enabled promotional content and spam. Reddit said at the time that these tools helped catch 23 million spam views per day and cut users' exposure to spam by about 20% in the period from January to March 2026, compared to the prior three months. Collins said that while some commercial spam is obvious to users, these tools were able to catch subtler instances, such as spam accounts that appeared to be people interacting with each other. Moderation is an ongoing Whac-A-Mole game, though, and as long as spammers keep finding workarounds, Reddit's work will continue. "The conversations that are happening on Reddit have value," Collins said. "We have to continue to adjust and continue to improve our efforts to remove that kind of content to make sure that it continues to be that great value of a place for users." What marketers are saying. Agency execs who spoke with CMO Insider said brands that use Reddit as intended - by listening to their consumers and being useful to them - won't be affected by the spam crackdown. "The brands that win will earn participation by being transparent, useful, and native to each community, while giving customers genuine reasons to talk about them," said Joe Gagliese, co-CEO of Viral Nation. Brian Yamada, chief innovation officer at VML, said the sugar rush some brands got by manipulating Reddit would be short-lived. Slop isn't going away, though, he added. "It's not just a Reddit problem," Yamada said. "The flood of generative AI slop is likely to continue as brands or companies are looking for shortcuts or cheats to get visible." Collins said Reddit has an appeals process to correct false positives and that it's letting suspected spam remain while it's under review. Moderators and users play an important gatekeeping role on Reddit, and Collins said that wouldn't change. Moderators can still review and approve content for their communities, but limiting suspected spam during a review period would help prevent users from being exposed to it, he said.

DA Davidson has lowered its price target on Reddit from $200 to $185 whilst maintaining a Buy rating, citing near-term risks to daily active user growth. Similarweb data shows continued pressure on global daily active users, with web traffic trends suggesting figures could fall below consensus estimates. However, Reddit's Q2 financial results paint a different picture. Daily active uniques rose 18% year-over-year to 130.3 million, whilst revenue reached $805 million, marking the eighth consecutive quarter with over 60% year-over-year revenue growth. Net income increased $164 million to $253 million. Despite strong financials, Reddit shares have fallen over 30% year-to-date. Of 35 analysts covering the stock, 66% rate it a Buy, with a median 12-month price target of $212.50 implying 34% upside.