About
Hi, I’m Bevan Stanely — most people call me Stanely.
I’m a GenAI Strategist & Engineer: I turn GenAI capability into business outcomes — and build the hard parts myself. I sit in the gap most companies struggle with right now: I can shape the AI strategy and build the hard systems underneath it — so when I say “this is possible” or “this won’t scale,” it’s earned, not hand-waved.
Right now I’m a Full Stack LLM Development Analyst on a Finance Data Transformation program for a major e-commerce client (via Accenture), owning requirements, source-to-target mapping, and PRDs for finance data-transformation use cases — carrying them through to the data layer, mapping finance data models onto the platform’s semantic layer, and building the SQL test harness that validates them — with Claude-assisted workflows in the loop. Before that I built production GenAI systems on the GenAI Experience Platform and shipped ML/CV pipelines end to end.
What I do#
I set the direction — and then I build the proof.
- AI strategy & solutions — scoping GenAI use cases, translating them into requirements and PRDs, and de-risking them by knowing exactly what the engineering costs. Backed by an MBA in Business Analytics (BITS Pilani) and IIBA ECBA certification.
- GenAI / LLM systems — RAG pipelines, multi-agent orchestration, and secure LLM integrations on AWS (Lambda, API Gateway, Bedrock) and GCP (Vertex AI, Gemini). Certified Databricks Generative AI and Azure AI Engineer Associate.
- The hard, low-level stuff most strategists can’t touch — GPU-accelerated speech-to-text in Rust (WhisperForge), quantized embeddings running in-browser via WebAssembly, and CUDA-parallelized optimization. This is my moat: I’ve actually shipped the difficult parts.
The bridge#
Most people in AI are one of two things: an engineer who can’t translate to the business, or a strategist who can’t tell buildable from fantasy. I’m deliberately both. My path — computational biology and biological sciences into GenAI engineering into finance-transformation strategy — is unusual on purpose. It means I can walk into a room, understand the business problem, and know within minutes whether the AI answer is a weekend prototype or a six-month platform.
Selected work#
- qrdrop — send a file straight from one device to another — no server and no signup: a QR code carries the key, the file goes over WebRTC, and nothing in between ever holds a readable copy. How it’s built · source.
- Quantized & Compressed Embeddings for Fast Edge Inference — running ML inference in the browser with Rust + WebAssembly (~4× smaller embeddings).
- Try It Live: Edge-Based Song Recommendation — a working in-browser WASM demo of the above.
- Mastering SQL Gaps and Islands — a practical data-analysis deep dive for inventory and subscription analytics.
- WhisperForge — GPU-accelerated Whisper reimplementation in Rust (Vulkan/DX12/Metal), published as a crates.io workspace.
Work with me#
I’m open to GenAI Strategist, AI Solutions Architect, AI Engineer, and Machine Learning Engineer roles where the mandate is to turn GenAI capability into business outcomes — and where being able to build the hard parts is an asset, not overkill.
The fastest way to reach me is LinkedIn — send me a message and I’ll reply.