Alex Zall — Agents manager

A node-based cockpit for AI video production. Commercials, series and film go from brief to final cut on one board: generation, review and montage side by side.

• →Node canvas where the brief and references feed every generation automatically
• →One catalogue of image and video models from fal, Runway, Higgsfield and ByteDance; studios bring their own keys, no markup
• →Long generation jobs run in Cloudflare Workflows; results land in R2 and reach the board live
• →Built with a team of coding agents: every ticket goes through a plan, review and tests in CI

An AI assistant that makes phone calls for you. You write the task in your language; it calls doctors, landlords and local businesses across Europe in theirs, and reports back with a comparison.

• →Two LangGraph agents: a wizard that turns a chat into a call plan, and an event-driven orchestrator that runs the calls, retries and follow-ups
• →Real phone calls through ElevenLabs voice agents over SIP, in 21 languages
• →Scheduling and retries on Cloud Tasks, live task state in Firestore, billing by the second
• →Prompts and models are checked with Evalite evals; every run is traced in LangSmith
• →Landing page optimised for SEO and GEO, so both Google and AI assistants find and cite it; mentions in AI search are tracked with Peec AI

• →Engineered an autoscaling GPU cluster on GCP for high-performance video rendering, reducing processing costs by ~40%.
• →Architected and deployed an AI-powered user feedback management system using LangGraph, automating categorization and sentiment analysis.
• →Implemented AI-driven SEO/GEO optimization strategies, increasing organic visibility.
• →Led a non-technical team of customer success and marketing specialists.

• →Developed core AI infrastructure including a high-performance system to intercept and analyze ChatGPT response streams in real-time.
• →Built an automated AI response categorization engine for deep analytics into LLM behavior.
• →Designed and managed a distributed network of parser bots using Playwright to extract data from non-API LLM interfaces.
• →Integrated AI analytics with Looker Studio for stakeholder dashboards.

Beeboard is saas project and process management solution

• →Adjusted team agile processes
• →Developed a technical vision and roadmap
• →Established coding standards and guidelines
• →Set up a CI/CD pipeline and deployment processes
• →Promoted learning culture within the team
• →Introduced monitoring and alerting systems
• →Improved database performance

Quickcommerce is a company behind Zapp. Part of the team including me later separated to Orcrist GmbH but continues to work with Zapp as well as other projects.

• →Developed an internal tool for warehouse and city managers called Zappboard. It aggregated data from every other system and presented it in an actionable way
• →Contributed to the last mile delivery app Zephyr. It showed the state of the order and rider's fleet on the map. And it was also deeply integrated with the Zappboard
• →Built tools for scanning and picking orders in the warehouse
• →Trained an ML model to predict the delivery time based on over 40 parameters and integrate it into the ordering process
• →Integrating with other storefronts like Uber Eats and Deliveroo
• →Contributed to the development of company-wide monitoring tools and ops infrastructure
• →I was the responsible person for organizing knowledge sharing in the company. I've organized and facilitated tech-talks and demos
• →Planned on-call rotations

MOIA is a company under the Volkswagen umbrella. The main challenge there was to optimize supply for the demand to match a particular van to a particular order and also re-distribute vans across the city to decrease waiting time. I have contributed to the following:

• →Built a real-time simulation service that simulated the vans' behavior based on real-life demand data
• →Contributed to a step-by-step simulation service. This service could simulate a week of real-life fleet behavior which allowed us to assess different ride-sharing algorithms
• →Contributed to a dispatching service that aggregated the state of the fleet in real-time showing the vans, their routes, and customers on the map. This service was the backbone that connected vans, customers, ride-optimization services, and fleet managers.

Zalando is Fashion e-commerce service. In my short time in Zalando, I've contributed to a few internal projects.

• →Product image labeling with AI. Service described the item by “looking” at its pictures.
• →Product image verification. Service checked that the item has all of the required images and that the images are in the right format. White background, shoes facing right and taking 75% of space, etc.

In my almost 5 years in SoundCloud I probably got the biggest skill boost in my career as I was working shoulder to shoulder with the most talented people from around the world. I've also learned how a healthy and pleasant work environment can look like. Some highlights from those years are

• →I am the man behind the canonical SoundCloud audio player on the web
• →I've also contributed to mobile audio players
• →In the copyright protection team, I developed a pipeline that identified and automatically took down any upload that infringed someone's copyright
• →I've learned scala Functional Programming techniques

YaTV is a pioneer TV streaming platform in Russia. I was responsible for the development of everything streaming related.

• →I've developed a video player
• →Streaming studio for content creators
• →Streaming backend to work with multiple CDNs