I design and build software. Not one or the other — both. Fractional CTO & Product Engineer based in NYC. I either build products solo end-to-end, or architect solutions and lead engineering teams through delivery.

Kolt's Table — A seasonal, plant-forward cooking platform designed and built solo. 1,500+ commits.

AI — user-facing features (Anthropic Claude):

• The Pantry — enter what's in your kitchen, Claude semantically normalizes and expands your ingredients, matches against the full recipe database, and returns ranked results with match percentage, match explanations, and a generated shopping list. Interactive homepage teaser drives traffic directly into the feature.
• Natural language recipe search — conversational queries ("something warming and quick for a weeknight") interpreted by Claude and translated into structured filters in real time
• Ingredient-aware search with synonym expansion — searches actual recipe ingredients, not just titles/descriptions. Synonym groups (spring onion = scallion = green onion, cilantro = coriander, + 10 more) ensure comprehensive results

Seasonal Intelligence Engine:

• Site-wide seasonal data config drives homepage, navigation, produce guides, and founder's note — all rotate automatically by month, zero manual updates
• "What's in Season" living reference page with interactive season tabs, clickable produce grid, and filtered recipe display
• Homepage redesign: seasonal hero, produce strip, curated recipes, editorial journal feature — 6 intentional sections, no redundancy

AI — Platform Intelligence (20 autonomous jobs, connected intelligence network):

• Cross-job causal intelligence — jobs trigger and inform each other through a dependency graph with tracked causal links (triggered, resolved, informed, validated). SEO audit triggers description polish; trending detection informs auto-featuring.
• Content lifecycle management — recipes tracked through published → performing → featured → declining → archived with 7-day vs 30-day engagement moving averages, daily EQS snapshots, and linear regression decay projection predicting unfeaturing dates
• Competitive intelligence — weekly Claude + web search scan of NYT Cooking, Bon Appétit, and Serious Eats for plant-forward content, cross-referenced against content library and zero-result search queries
• Engagement Quality Score (EQS) — proprietary composite of ratings, analytics, favorites, scroll depth, time on page, and seasonal context with governance rules and one-click revert
• Newsletter drafting with Mailchimp integration — Claude drafts editorial newsletters, Apply creates a Mailchimp campaign ready for review and send
• Internal linking with editorial merge — AI identifies thematic connections between content, Apply writes to junction table, links merge into "More to Explore" sections alongside algorithmic pairings
• Confidence scoring — Levenshtein-based change magnitude scoring on all AI-proposed edits with recently-edited penalty, performance adjustments, SERP preview mockups
• Real-time SSE event stream — live activity indicators on the admin dashboard via Server-Sent Events with auto-reconnect
• Intelligence briefing — daily Claude-generated 3-sentence platform briefing with composite health score (6 weighted dimensions) and vital signs strip
• Weekly intelligence report — health trend, lifecycle distribution, EQS movers, job activity breakdown, causal chain count, approval throughput
• 24-hour radial clock — SVG visualization of job schedules and execution history with color-coded legend
• Image generation review — Gemini-generated hero and seasonal images with full-screen review modal (Download/Regenerate/Dismiss)
• Anonymous engagement analytics pipeline (zero PII) — time on page, scroll depth, view counts
• Content analysis — recipe description polish, meta description optimization with SERP preview
• Content and pantry gap analysis — identifies missing recipes based on seasonal produce and site data
• SEO three-stage pipeline: automated audit → Claude proposes editorial fixes → approval queue with side-by-side comparison
• Codebase health scanner — dependency audit, security vulnerabilities, code quality metrics, database health
• Trending detection, draft monitoring, and cron scheduler with live sync

Governance & Approval System:

• Centralized review queue for all AI-proposed changes with keyboard navigation
• Side-by-side current vs proposed comparison with confidence bars
• Apply writes to database only after human approval, full audit trail
• Newsletter → Mailchimp, SEO fixes → recipe DB, internal links → junction table, descriptions → recipe DB
• Jobs without functional apply gracefully acknowledge without error

AI — Recipe Entry (two-stage):

• Paste raw recipe text → Sonnet parses ingredients/instructions in ~2s → Opus classifies across 100+ filter categories
• Strict classification rules: only selects what's true as written, never speculative

Platform:

• 180KB production bundle (53% smaller than NYT Cooking)
• Sub-50ms recipe search via PostgreSQL GIN indexes
• 96/100 Lighthouse performance scores across all metrics
• 60+ normalized PostgreSQL tables, zero technical debt
• User dashboard with engagement stats, searchable favorites, dietary preferences
• Full CI/CD pipeline — Docker, GitHub Actions, zero-downtime deploys
• New Relic APM, Redis caching, Elasticsearch, Kibana

env-doctor — Zero-dependency CLI that diagnoses environment variable issues in any Node.js project.

• Scans source code for process.env usage
• Compares against .env.example — surfaces missing, undocumented, and unused vars
• Exit code 1 on issues, making it CI-pipeline friendly

• Portfolio: koltadams.com
• Email: kolt@koltadams.com
• LinkedIn: linkedin.com/in/koltadams