Reliable AI for real work.

Led a team of 2 researchers and data scientists in a small language model research program for a multi-billion-dollar consumer technology company, post-training sub-500M parameter models across multiple model families with reinforcement learning (GRPO, RLVR) for math reasoning (GSM8K) and tool calling, over two phases of 60 and 55 days spanning 6 months.

Designed an enterprise AI reference architecture for a 5,000+ employee organization, covering governance and responsible AI, monitoring and audit logging, authentication and authorization, model deployment, MCP integration, and the data layer required to deliver secure AI solutions at scale.

Compared large and small language models and AI platforms on quality, cost, latency, and privacy to recommend right-sized solutions rather than defaulting to the largest model.

Built and deployed AI agents for clients, including a document agent specialized in understanding and analyzing documents, and ran its prompt injection security assessment ahead of full rollout, writing the mitigation report that got it approved for release.

Took over a stalled client automation project that had no evaluation process and conflicting rules, working with a junior colleague to rebuild the pipeline, add error analysis, and get the client relationship back on track.

Delivered a production order recommendation and demand forecasting system using probabilistic and Poisson modeling, picking up a stalled implementation from another developer and reaching a client approved handover.

Worked with a junior colleague to build a demand forecasting engine (ARIMA, SARIMA, Prophet) for a manufacturing and logistics client, modeling at daily product level to improve on a single aggregate model.

Ran 2 knowledge-sharing sessions with client teams on understanding and adopting AI.

Across PoC, MVP, and production engagements, framed client problems, assessed feasibility, value, and risk, wrote business cases, and recommended whether to scale, adapt, or stop based on evaluation evidence.

Supported proposal and delivery work with enterprise clients across building materials, industrial automation, legal technology, and logistics, alongside account and leadership teams handling stakeholder management and go/no-go decisions.

Worked with product and leadership teams at a document intelligence platform used by law firms and private equity funds, presenting user behavior findings to the CTO, CPO, and product leads that helped shape 3 later AI initiatives.

Designed, organized, and presented more than 10 internal and client knowledge-sharing sessions on software engineering principles, AI agents, adoption, and the foundations needed for practical use. Each session combined a concise theory section, a live demo, and a purpose-built hands-on exercise.

Organized and delivered a public, hands-on event on AI agent harness engineering at the AI Tech Campus in Eindhoven.

Helped drive adoption of AI coding agents across the team, scaling to 6 agents running in parallel, and co-authored an internal maturity survey and starter kit.

Built a working internal AI meeting note taker prototype in 7 days, covering speech to text, LLM summarization, and Azure deployment, reaching 65 to 75% completion in the first sprint.

Took part in 30 or more technical interviews and mentored 3 junior and medior engineers on live client projects.

Co-designed a 6-week onboarding program across technical, business, and non-technical tracks, used for 10 or more new hires across 2 offices.

Maintained the company website (Next.js, Payload CMS, MongoDB, Vercel, Azure), improving technical SEO to a 99 Lighthouse score and building a drag and drop CMS editor that replaced a paid vendor tool.