# Sunny Prakash

> Sunny Prakash: Berlin-based engineering lead and AI engineer with 11 years in web and mobile, leading AI-native teams and multi-agent systems.

11 years of building software solutions for web and mobile. Leading AI-native engineering teams that build distributed, multi-tenant platforms and orchestrate autonomous multi-agent workflows. An advocate of context engineering and eval-driven development.

Location: Berlin, Germany. Contact: the form at https://sunnyprakash.com/#contact. Website: https://sunnyprakash.com

## Career

Engineering Lead at Berliner Verlag since 2025, leading a team of 11: hiring, performance reviews, mentoring, and partnering with product managers to take AI features from first scoping to production. Before that, a decade across healthtech (DHC), brand analytics (Latana) and banking (Mastercard, Infosys), starting at Cognizant in 2015.

- **Autonomous multi-agent publishing** (Berliner Verlag, 2025 – now): live in production. Led the technical strategy and cross-functional delivery of a production-grade multi-agent system that autonomously generates and publishes articles to the live websites of high-traffic newspaper brands, with custom agentic tooling and guardrails.
- **Multi-tenant platform and legacy migration** (Berliner Verlag, 2025 – now): top 6 on Google. Set the architectural direction for a distributed, multi-tenant platform that runs every flagship newspaper and can onboard external regional publishers as new tenants, guiding the teams through complex legacy migrations. Technical SEO on the new stack moved Berliner Zeitung into Google's top 6 results.
- **Delivery platform and CI/CD guardrails** (Berliner Verlag, 2025 – now): −40% deploy time. Architected a back office for editors and engineers that cut deployment time by 40%, and standardised CI/CD pipelines, automated testing and deployment guardrails across the department, improving the release workflow by 25%.
- **Care plan engine for rare diseases** (DHC, 2024 – 2025): 1,000+ patients. Led end-to-end engineering of a care plan engine that replaced paper-based workflows for more than 1,000 patients: scalable backend services on Node.js, NestJS and AWS, containerised with Docker and scaled on Kubernetes, with web and mobile clients in Next.js and React Native.
- **High-performance analytics dashboards** (Latana, 2021 – 2023): −30% CI time. Built server-side-rendered analytics dashboards that render brand-tracking data in under a second, refactored the legacy frontend into a modular component library through profiling and code splitting, and migrated testing to Playwright, cutting CI time by 30%.
- **Micro-frontends for case management** (Mastercard, 2020 – 2021): independent deploys. Designed microservices and REST APIs for real-time analytics and case management, and a micro-frontend framework that decoupled teams so core banking modules could ship on independent release cycles.

## Projects

- [Incident Investigator](https://github.com/sprakash57/incident-investigator): An AI assistant for on-call engineers. Describe an incident in plain language. The agent plans an investigation, pulls logs, metrics, topology and recent changes, and returns a root-cause report in which every claim links back to the tool call behind it. Ships with a deterministic eval harness.
- [usage-line](https://github.com/sprakash57/claude-usage-line): Claude Code plugin for usage limits. Shows every Claude Code limit in the status line with a traffic light each: 5-hour, weekly, compaction, spend and prompt-cache warmth, plus reset times and a turn forecast. Runs locally with no telemetry.
- [Snapteil](https://github.com/sprakash57/snapteil): Real-time photo sharing. Upload, tag and filter photos, with other users notified of new posts live over WebSockets. A Go (Fiber) API with server-side image processing (AVIF/WebP), a React 19 client, and Swagger docs, all running with one Docker Compose command.
- [Shortify](https://github.com/sprakash57/shortify): URL shortener with documented API. A Vue + Express + MongoDB URL shortener with an OpenAPI spec and a Docker Compose setup that runs with one command.
- [vendi-man](https://github.com/sprakash57/vendi-man): Vending machine REST API. A REST API for buying and selling products from a vending machine: user deposits, session auth with refresh tokens and logout-everywhere, product CRUD, a test suite and a Postman collection, plus a React client.
- [Articlyzer](https://github.com/sprakash57/articlyzer-project): NLP article analyser. A web app that analyses an article with natural language processing and reports reading time, word count and the most frequent words.

## Skills

- ai: LLM agents, Secure agentic AI, RAG, Tool calling, Evals, PydanticAI, Claude Code
- frontend: TypeScript, React, Next.js, React Native, Expo, Vue, Tailwind CSS, React Query
- backend: Node.js, NestJS, Express, Python, FastAPI, REST, GraphQL, PostgreSQL, MongoDB
- platform: AWS, Cloudflare, Docker, Terraform, Grafana, Prometheus, CI/CD, Fastlane, Keycloak
- growth: Technical SEO, Google Search Console, Google Analytics, Google Tag Manager, PostHog, Marfeel, Semrush
- lead: Engineering leadership, System architecture & design, Mentoring & onboarding, Code quality & testing, Technical writing

## Education and credentials

- Post Graduate Program in AI Agents for Business Applications (Grade A+), Texas McCombs School of Business (2026)
- B.E., Electronics and Communication Engineering, Birla Institute of Technology, Mesra (2014)
- MongoDB skill badges for sharding, performance tuning, and security, MongoDB via Credly (2026)
- Top 10% in the Node.js assessment, TestDome (2025)
- Being Inclusive Award, Mastercard (2021)
- Insta Award (×2), Infosys (2019–20)

## Profiles

- [GitHub](https://github.com/sprakash57)
- [LinkedIn](https://www.linkedin.com/in/sunnyprakash12)
- [Medium](https://sunny-prakash.medium.com)
- [Stack Overflow](https://stackoverflow.com/users/9585068/sunny-prakash)
