> Markdown version of [/videos/2161-ai-vector-search-at-scale-ewa-szyszka-ewa-szyszka?t=298](https://www.wearedevelopers.com/videos/2161-ai-vector-search-at-scale-ewa-szyszka-ewa-szyszka?t=298). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # AI Vector Search at Scale - Ewa Szyszka - Ewa Szyszka Ewa Szyszka argues the strongest AI founders combine technical fluency with deep domain expertise. Learn to scale vector search, slash LLM costs, and deploy robust agentic evaluations. - **Speakers:** Ewa Szyszka - **Event:** Coffee With Developers - **Published:** August 31, 2026 - **Duration:** 11:14 - **URL:** https://www.wearedevelopers.com/videos/2161-ai-vector-search-at-scale-ewa-szyszka-ewa-szyszka ## Summary Managing large language model costs requires precise data retrieval rather than overwhelming context windows with unstructured information. By evaluating, chunking, and embedding data before storing it in a vector database like Qdrant, developers can maximize data utility while minimizing token usage and avoiding strict rate limits. Efficient search algorithms, such as HNSW and TurboQuant, enable rapid processing of complex, multimodal datasets at scale, ranging from video transcripts and audio parsing to IoT edge environments. Moving from pristine demo datasets to real-world deployment reveals the necessity of rigorous data cleanup and continuous testing. Implementing agentic evals ensures that deployed AI tools and skills are perpetually evaluated against datasets containing perfect golden answers. This continuous evaluation cycle is crucial for handling messy outliers, building solutions for anomaly detection in unstructured formats like CCTV feeds, and maintaining robust production systems after initial deployment. The rapidly evolving AI landscape shifts the startup founder advantage from sheer technical capability to deep, specialized knowledge. While modern agentic development tools lower the barrier to entry, the true competitive moat lies in combining technical fluency with specific industry knowledge to solve pressing, monetizable problems. Because navigating the bleeding edge requires continuous learning—echoing the core demands of a developer relations engineering role—successful founders must realize they cannot master every new model. As Ewa notes, "domain expertise plus technical skills is where I think strongest founders lie," ultimately requiring them to delegate and build teams of highly obsessed specialists rather than attempting to be jacks-of-all-trades. **Keywords:** vector search optimization, token usage minimization, LLM context window management, multimodal data retrieval, qdrant vector database, HNSW algorithm, turboquant processing, edge environment IoT, unstructured data parsing, anomaly detection models, agentic evals, continuous deployment evaluation, golden answers datasets, developer relations engineering, AI startup founding, agentic software development, technical founder moat, AI domain expertise ## Chapters 1. **Optimizing token usage with vector search retrieval** (00:00) — Efficient information retrieval using vector databases prevents hitting rate limits and reduces token costs. 1. **Preparing and chunking data for vector databases** (01:23) — Understanding use cases and properly chunking data before embedding ensures highly relevant search results. 1. **Scaling search and managing multimodal data types** (02:35) — Vector databases handle large-scale multimodal data like audio and video using algorithms like HNSW and TurboQuant. 1. **Bridging customer needs and technical solutions in developer relations** (03:55) — Developer relations engineers build practical workshops based on real-world customer stories and ecosystem demands. 1. **Exploring multimodal AI applications beyond standard language models** (04:58) — Processing unstructured data like CCTV footage and audio opens new possibilities for anomaly detection and advanced analytics. 1. **Managing dirty data and continuous agentic evaluations** (06:06) — Implementing continuous evaluation processes with golden answer datasets is critical for handling real-world data errors in production. 1. **Comparing startup founder experiences with developer relations roles** (07:28) — Building in public as a founder or advocate sharpens technical skills and fosters rapid continuous learning. 1. **Combining domain expertise with technical skills for AI startups** (08:40) — Founding successful AI companies requires pairing technical fluency with deep domain knowledge and delegating specialized tasks. ## Related Moments - [Evaluating venture capital value and AI sustainability](https://www.wearedevelopers.com/videos/100104-from-saas-to-space-the-defensible-value-vcs-are-looking-for) (from "From SaaS to Space: The Defensible Value VCs Are Looking For") - [Evolving venture capital strategies amid artificial intelligence democratization](https://www.wearedevelopers.com/videos/100049-beyond-the-wrapper-technical-bets-that-vcs-back) (from "Beyond the Wrapper: Technical Bets That VCs Back") - [Introduction to generative AI and vector search](https://www.wearedevelopers.com/videos/844-enter-the-brave-new-world-of-genai-with-vector-search) (from "Enter the Brave New World of GenAI with Vector Search") - [Audience questions on data integration and future action models](https://www.wearedevelopers.com/videos/899-creating-industry-ready-solutions-with-llm-models) (from "Creating Industry ready solutions with LLM Models") - [Essential AI and human skills for future teams](https://www.wearedevelopers.com/videos/1623-breaking-silos-successful-collaboration-between-tech-business-teams-in-complex-enterprise-systems) (from "Breaking Silos: Successful Collaboration Between Tech & Business Teams in Complex Enterprise Systems") - [Dismantling AI hype and focusing on practical use cases](https://www.wearedevelopers.com/videos/1356-from-learning-to-leading-why-hr-needs-a-chatgpt-license) (from "From Learning to Leading: Why HR Needs a ChatGPT License") ## Related Articles - [Navigating the AI Shift](https://www.wearedevelopers.com/magazine/629-navigating-the-ai-shift) - [SEO in an AI world - Google vs. ChatGPT and survival tips for content creators](https://www.wearedevelopers.com/magazine/534-seo-in-an-ai-world-google-vs-chatgpt-and-survival-tips-for-content-creators) - [Stephan Gillich - Bringing AI Everywhere](https://www.wearedevelopers.com/magazine/489-stephan-gillich-bringing-ai-everywhere) - [Exploring AI: Opportunities and Risks for Developers](https://www.wearedevelopers.com/magazine/522-exploring-ai-opportunities-and-risks-for-developers) ## Related Jobs - [Principal Engineer - AI Search & Vector Infrastructure](https://www.wearedevelopers.com/jobs/ext/319507-principal-engineer-ai-search-vector-infrastructure) at **Redis** - [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat** - [Principal Engineer - AI Search & Vector Infrastructure](https://www.wearedevelopers.com/jobs/ext/353953-principal-engineer-ai-search-vector-infrastructure) at **Redis** - [Principal Engineer - AI Search & Vector Infrastructure](https://www.wearedevelopers.com/jobs/ext/381484-principal-engineer-ai-search-vector-infrastructure) at **Redis** - [Partner Sales Director - AI Alliances - Model Providers](https://www.wearedevelopers.com/jobs/48429-partner-sales-director-ai-alliances-model-providers) at **Dynatrace** - [Principal Software Engineer, Enterprise AI Platform](https://www.wearedevelopers.com/jobs/ext/1467292-principal-software-engineer-enterprise-ai-platform) at **GitHub**