Technologist shaping systems where design craft and automation meet.
I’m Dan Becker—an engineer with a BFA in graphic design, building products, tooling, and processes that keep teams moving. From design system governance to predictive pricing and AI-assisted workflows, I focus on practical systems that feel considered.
What collaborators say.
Dan is a highly collaborative developer who enjoys interacting with others and sharing ideas. He's talented with both engineering and design, and this allows him to execute with a high level of thoughtfulness and consistent excellence. He shows an enthusiasm for web development that is refreshing to work with. He's a great communicator with a friendly personality, and would be an excellent leader.
Amelia Y.Design & engineering impact.
of professional software engineering experience spanning process automation, image generation, design tooling, full stack web development, and enterprise UI design systems.
of professional design experience across home decor, branding, print, UI/UX, and digital product design.
of professional design & engineering experience creating impact.
Citrus Design System @ Kroger | skillhubai.com | DesignerPower UXP | MyPrice
Design sensibility meets systems thinking.
I'm a software engineer with a BFA in graphic design—pairing visual systems thinking with practical automation. After years leading design and merchandising tooling inside a home décor company, I now focus on platform and design system work with an eye toward AI-driven workflows.
Recent projects include an AI development certificate with Python and hands-on experiments in deep learning and computer vision. I'm currently pursuing graduate study in AI to deepen that practice.
Curiosity drives everything I build. Whether it's a data-backed pricing model, a design operations pipeline, or an experimental creative tool, I'm interested in the systems that keep teams moving.
Outside of engineering I've been playing guitar for 25 years and producing ambient music for 14 — because iteration, whether in code or sound, is where the ideas surface.
Product, automation, and design system engagements.
Explore projects spanning enterprise design systems, applied AI, and automation. Filter by focus area or adjust the timeline view to dig deeper.
SkillHub AI
Situation: Hiring teams lose high-signal candidates to brittle keyword filters while job seekers lack tools to surface meaningful career context. Traditional resumes favor prestige over potential, and existing resume builders force users into rigid templates or rely purely on conversational AI without structure or version control.
Task: Build dual-platform solution: (1) AI-native semantic profiling system with MCP server for intelligent candidate matching, and (2) Intuitive resume management platform that bridges the gap until enterprises adopt semantic profiles—combining fine-grained editing, AI-powered enhancements, versioning, and automated tailoring.
- MCP Server Architecture: Built Model Context Protocol server exposing three tools (search_candidates, get_candidate_profile, analyze_job_fit) enabling any LLM-powered system (Claude Desktop, custom AI agents, recruiting platforms) to perform natural language candidate searches with semantic understanding beyond simple RAG.
- Multi-LLM Orchestration: Architected 5+ semantic search strategies (simple, enhanced, conversational, streaming, AI-first, scalable) with Claude + Gemini coordination, fallback handling, and rate limiting for production-grade reliability.
- Resume Management Platform: Engineered full-featured resume builder with tabbed interface for fine-grained section editing, direct field modification, AI-powered enhancement suggestions, and LLM-assisted revision requests—moving beyond pure chatbot interactions to structured, visual editing.
- Artifact System: Implemented versioning and variants allowing users to maintain canonical resume while generating tailored versions for specific jobs, with automated job-description matching and customization workflows.
- AI Review & Enhancement: Built "AI Review" feature that analyzes resumes against industry standards, generates targeted improvements, creates tailored versions automatically, and provides actionable feedback—integrated throughout the editing experience.
- Semantic Profile Generation: Developed embedding pipeline and skills graph parsing resumes into competencies, impact signals, and trait vectors (e.g., "Systems Thinker," "Full-Stack Innovator") with experience-level calculations and role progression analysis.
- Production Infrastructure: Full authentication system, MongoDB with PII encryption, background processing with Gemini/Claude APIs, analytics instrumentation, comprehensive testing (Vitest, snapshot tests), PM2/Docker deployment.
- Prototyping Process: Started as candidate assessment POCs for EmployerTube.com, tested with real hiring managers, integrated successful features into production platform, then spun off SkillHub as isolated environment for experimental capabilities (MCP, semantic profiling, agentic workflows)
Result: Launched dual-platform beta: (1) MCP server enabling LLM-powered recruiting agents with superior candidate data quality compared to pure RAG approaches, currently accessible via Claude Desktop and integrable into custom AI systems; (2) Resume management platform offering job seekers intuitive tools for creating, editing, enhancing, and tailoring resumes with AI assistance—supporting immediate needs while semantic profiles gain enterprise adoption. Live with pilot companies, instrumented with analytics loops, prepared to sunset if traction doesn't materialize.
TechFitPath career evaluator
Situation: Multidisciplinary technologists often struggle to describe whether they are architects, design technologists, deep engineers, or platform builders, which slows career planning conversations.
Task: Build a lightweight evaluator that collects qualitative and quantitative signals, persists the data, and returns an interpretable recommendation instantly.
- Scripted identity prompts, Likert scales, and capability toggles tied to a React state model that powers Recharts radar visualisations and insight cards.
- Implemented an App Router API that stores every submission in Vercel Postgres and calls the OpenAI Responses API to calculate cluster weights, traits, and rationale.
- Added JSON export, reset, and accessible micro-interactions so individuals and coaches can reuse the insights in their own systems.
Result: Launched TechFitPath.com as a shareable tool; early cohorts map their archetypes in minutes and walk away with artefacts they can bring into mentorship or staffing discussions.
DesignerPower UXP
Situation: Designers exported assets from Photoshop, then manually uploaded to Widen DAM, causing delays and inconsistent metadata.
Task: Create an in-app workflow that keeps designers focused in Photoshop while ensuring assets land in DAM with clean metadata.
- Authored modular UXP panels that batch production prep, compliance checks, and DAM delivery without leaving Photoshop.
- Integrated Widen APIs plus Anthropic Claude to generate keywords, categories, and descriptions that respect controlled vocabularies.
- Instrumented usage analytics and guardrails so product leadership could monitor adoption and metadata quality.
Result: Drove measurable increases in DAM adoption, improved metadata consistency, and cut asset hand-off time by keeping creative work in one surface.
Citrus live theming lab
Live theming lab showing citrus tokens across product surfaces.
Situation: Kroger's product teams and leadership needed to feel how citrus tokens and interaction patterns behaved at scale, but static docs and slide decks failed to convey the system's nuance.
Task: Build a memorable demo that visualizes theming and design system innovations through interactive exploration.
- Drafted scripted voice over with LLMs and used WellSaid to generate AI-narrated walkthroughs.
- Recorded high-fidelity video captures of theming transitions, accessibility modes, and motion patterns for embedding in presentations.
- Combined the narration audio and video clips into a 2 minute highlight reel that could be shared asynchronously.
Result: The demo became a fixture in quarterly reviews and onboarding sessions, helping secure budget for citrus adoption and giving teams a visceral reference for the system's intent.
MyPrice Agent: AI-powered product intelligence
Situation: A merchandising team relied on slow factory responses for pricing, delaying quotes and leaving sales reps without timely guidance.
Task: Deliver an internal tool that predicts per-unit manufacturing cost with accuracy high enough to replace manual quoting.
- Cleansed 120k+ historical quotes, normalising units, repairing categorical data, and vectorising mixed inputs for modelling.
- Built and tuned TensorFlow regression pipelines with one-hot, TF-IDF, and log-transformed targets to stabilise predictions.
- Wrapped the model in a lightweight service and reporting UI so merchandisers could query scenarios instantly.
Result: Achieved 1.45% MAPE in testing, projected to return 1,300+ hours of analyst time annually while giving sales real-time pricing guardrails.
Soulfolio Co
Situation: A gifting startup needed customers, designers, and production staff to collaborate on personalised products without emailing static proofs.
Task: Launch a web platform that supports shared curation, lightweight editing, and automated production file delivery.
- Built collaborative project spaces with real-time status so customers and internal teams tracked decisions together.
- Implemented a browser-based image editor for applying templates, colourways, and personalised text on top of library assets.
- Automated the backend to bundle approved designs into print-ready files as soon as an order closed.
Result: Replaced a fragmented proofing process with a single flow from concept to production, unlocking faster turnaround for bespoke gifts.
Opens a full-screen demo of the personalized image editor built for Soulfolio.
MyPrice
Situation: Category managers hunted for historical pricing across shared drives, manually copying quotes into ad-hoc spreadsheets.
Task: Centralise quoting history behind search and comparison tools so sales could respond to customer requests quickly.
- Wrote network crawlers that parsed disparate quote formats and normalised data into a MongoDB catalogue.
- Designed a basket workflow that let reps curate multiple quotes, compare deltas, and export client-ready packets.
- Added permissions and caching so the tool scaled across teams without overloading shared storage.
Result: Cut hours of manual digging per request and gave sales a single source of truth for pricing conversations.
Dapper
Situation: Marketing teams needed refreshed colour themes for banner apps, but changes required developer intervention and long release cycles.
Task: Prototype a governed theming tool that let non-developers propose colour tweaks while keeping the design system in sync.
- Modelled design token dependencies so updates to base colour ramps cascaded predictably across components.
- Built a React interface with guardrails and previews, enabling stakeholders to experiment without touching code.
- Integrated with build tooling to export approved palettes for downstream apps.
Result: Demonstrated how the design system could support rapid theming requests and informed the roadmap for managed colour customization.
Looksee
Situation: Merchandisers validated new artwork by emailing static PDFs to select customers, collecting unstructured feedback days later.
Task: Provide a lightweight feedback portal that captures buyer sentiment in real time and aggregates insights for the studio.
- Built curated collections with shareable links so buyers reviewed art in a focused, branded environment.
- Implemented quick reactions (hearts, comments, binary votes) and live dashboards for merchandising to monitor sessions.
- Iterated on the UI—from swipe-style prototypes to the final grid layout—based on buyer testing.
Result: Compressed feedback cycles from days to minutes and gave merchandisers data-backed confidence on which designs to advance.
Dr. Prepper
Situation: Production artists manually resized art, created mockups, and generated print files—work that bottlenecked merchandising launches.
Task: Automate artwork ingestion and rendering so non-developers could trigger production-ready assets on demand.
- Built ExtendScript pipelines that validated incoming art, mapped it to product templates, and generated multi-surface renders.
- Provided a lightweight React control panel and drop-folder workflow so designers and sales could kick off jobs themselves.
- Implemented self-healing routines that retried failed renders and reported status through dashboards.
Result: Increased design throughput by 477% and enabled near-immediate availability of e-commerce imagery for 31 product categories.
META — Microcontroller Enhanced Tablet Accessory
Situation: Surface Pro users lacked a travel-friendly way to trigger complex shortcuts without carrying full keyboards or remotes.
Task: Invent and validate a hardware accessory that adds programmable buttons to the tablet while fitting premium industrial design expectations.
- Designed industrial concepts, 3D printed housings, and embedded microcontrollers to map buttons to user-defined shortcuts.
- Filed patents and trademarks, built firmware in C++/Arduino, and produced a polished Kickstarter pitch with video and launch assets.
- Coordinated with manufacturers to scope tooling, BOM, and fulfilment options ahead of campaign launch.
Result: Produced a production-ready prototype and Kickstarter campaign that demonstrated market appetite for customizable tablet controls.
The Code Wheel Project
Situation: Traditional QR codes carried limited data, constraining interactive print experiences the studio wanted to explore.
Task: Prototype an encoding scheme that increases data density without requiring specialized scanners.
- Created a colour-encoded data model translating binary pairs into a constrained palette for circular codes.
- Developed proof-of-concept generators and reading routines that demonstrated compatibility with existing imaging hardware.
- Packaged the concept for a prospective Kickstarter with visual storytelling and technical documentation.
Result: Showcased a viable path to 4x data density within the same surface area, opening conversations about enhanced packaging interactions.
