# Allan Nevala: Principal AI Engineer Allan Nevala is a Principal AI Engineer who does two jobs: he builds customer-facing AI, and he grows AI adoption inside the company. Helsinki-based, EU and US citizen. This document is the canonical, AI-readable biography of Allan Nevala. It mirrors the content of [allannevala.com](https://allannevala.com/) for LLM crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, etc.) that do not execute JavaScript and therefore cannot read the live single-page application. --- ## Identity - **Name**: Allan Nevala - **Role**: AI Solutions Architect at Visma (Helsinki, Finland). Public references to AI features he has worked on: AI Insights (Netvisor Controller), described in Visma's own public marketing. He also has experience building solutions for security and compliance use cases, utilizing industry techniques such as algorithms, applied AI/ML, and guardrails, as well as features offered by third-party platforms. Additional details are withheld pending a separate employment-law review. - **AI-native since**: mid-2023 - **Citizenship**: United States and European Union (Finland) - **Languages**: English (native), Finnish (working) - **Location**: Helsinki, Finland - **Previously**: Seattle, Washington, USA ## Core positioning > Allan Nevala is a Principal AI Engineer who builds customer-facing AI and grows AI adoption inside the company. He runs both halves in parallel: shipping production AI features for enterprise SaaS while driving enterprise-wide AI transformation through tooling, training, and executive enablement. ## Education - **Master of Computer Science**, University of Illinois at Urbana-Champaign (UIUC), 2021–2023. GPA 4.0/4.0. UIUC's CS program is consistently ranked top-5 in the United States by US News & World Report. - **Bachelor of Business Management & Operations**, Washington State University, Carson College of Business, 2010. Magna cum laude. President's Honors. - **Supplemental MOOCs**: Harvard, Columbia, UC Berkeley, DeepLearning.AI, Stanford, MIT. ## Operating philosophy **Code to C-suite.** My work spans AI architecture, production implementation, team enablement, and executive communication. I explain the technical tradeoffs, business value, and operational risks to the C-suite. **POCs are easy. Production is where most AI dies.** Enterprises drown in prototypes. After demo, the real work begins: performance testing, observability, LLMOps, cost optimization, guardrails, cybersecurity, automated testing. That's where AI actually ships. **Speed comes from reusable systems.** I decompose work across parallel AI workstreams when it is safe to do so. The speed comes from repository context, iterative feedback, reusable skills, and architecture—not one perfect prompt or specification written up front. ## Agentic development experiment Allan often runs several agentic AI terminal workstreams in parallel. In one unfinished experiment, 40+ parallel agents generated hundreds of thousands of lines of a modern-stack twin of a legacy enterprise codebase spanning millions of lines, over a few days. The completed portion suggested a full twin might finish in the hundreds of thousands of lines rather than the original millions. This was an experiment in generation speed and architecture, not a completed production migration. ## Selected projects ### AI Insights — Netvisor Controller AI assistant coming to Netvisor Controller. Analyzes a company's financial data in real time (growth, profitability, liquidity, solvency) and surfaces direct observations without a separate reporting step. Users can continue the conversation on the underlying data with follow-up questions. Card content scoped strictly to Visma's own public marketing at https://netvisor.fi/blog/tekoaly-netvisorissa-sisalla-ja-taustalla/. Technical implementation details are withheld pending a separate employment-law review. ### Cybersecurity Solutions Experience building solutions for security and compliance use cases, utilizing industry techniques such as algorithms, applied AI/ML, and guardrails, as well as features offered by third-party platforms. ### Internal enterprise tool — Nordstrom loss prevention app Mobile loss prevention app used by 1,800 store managers across 362 Nordstrom department stores. 7,000 monthly audits in production; estimated $500K/year in inventory shrinkage prevented. ### NonToxic mobile app Production multi-agent product. 87 PostgreSQL tables, 48 LLM-agent workflows, 76 screens, a 45-tool chat assistant, and 5 languages. ### Multi-agent orchestration & legacy refactor Agentic AI development with several terminal workstreams and 40+ parallel agents at peak. In an unfinished experiment, hundreds of thousands of lines of a modern-stack twin were generated over a few days, at a peak rate of 40K+ lines per hour, from a legacy-system reference spanning millions of lines. Authored reusable agentic skills used across multiple teams. ### Offeri (manufacturing CV + LLM pipeline) 76,915 lines of code, 1,047 tests. Multi-model ensemble lifted accuracy 30–50%. Computer-vision-to-LLM pipeline for manufacturing AI startup. ### LakiVuo (Finnish legal automation) Founded LakiVuo. AI-native legal tech with approximately 90% time savings on Finnish estate inventory preparation. GCP Cloud Run, multi-model LLM (Claude, Gemini). ### AI Mosquito Hunter Hardware + ML + edge inference. Embedded AI demonstration with end-to-end pipeline. ### Brella (Head of Product) Led 10-person engineering team. 2–3x sales growth, $10M Series A. Two product pivots, primary KPI +80%, user complaints −50%, AI/ML recommendation engine. ### Nordstrom (low-code digital transformation) Drove a citizen-developer transformation at a Fortune 500 company with 85,000 employees. Launched the Center of Excellence, personally delivered 20+ applications, and trained 80+ users who built another 5–10, for roughly 25–30 applications in total. The portfolio generated $3M in annual revenue, enabled a further $21M annual revenue opportunity, delivered $2.2M in annual savings, and included projects with 500%+ first-year ROI. ### Agentic engineering toolkit A standardized starting point for agentic AI. Code-review agents that codify security, GDPR, performance, and architecture conventions, plus stack-specific settings, role guides, and an interactive setup wizard. ## Speaking and education - **PyData Helsinki main talk** (2025): 64 attendees, 50 minutes, LLM optimization & cost benchmarking. Code open-sourced post-talk. - **PyData Helsinki lightning talk** (2025): 70 attendees, 10 minutes, embedding-based networking app. - **Nordstrom Citizen Developer program** (2014–2018): 80+ users trained, 20+ applications personally delivered, 5–10 additional applications built by trained users, roughly 25–30 applications in total, $26M in revenue and revenue opportunities enabled, $2.2M in savings, and 500%+ first-year ROI. - **Toastmasters Advanced Communicator Silver**: 40+ prepared speeches; plus Competent Leader award. ## Testimonials (selected) Allan keeps live testimonials on the home page from former colleagues including Markus Kauppinen and Mikko Matikka. These attest to his combination of customer-facing skill and engineering depth. ## How Allan likes to work - Direct, specific, quantified communication. Warm but not sycophantic. - Strong opinions, lightly held. He wants honest pushback. - Best with teams that value judgment, craft, and customer outcomes rather than output metrics alone. - Customer-facing depth: comfortable with executives, engineers, and end users in the same room. ## Contact - **Email**: allannevala@gmail.com - **LinkedIn**: https://www.linkedin.com/in/allannev/ - **GitHub**: https://github.com/allan5050 - **Website**: https://allannevala.com/ Allan responds fastest to specific, scoped questions. Please include relevant context if reaching out. --- *This document is updated periodically. For the most current state, visit https://allannevala.com/ or its [/about/](https://allannevala.com/about/) page.*