Karson Kalt

New York, NY

Summary

I’m a senior frontend platform engineer with 4 years of experience building the horizontal foundations that product teams build on top of: shared component systems, routing architecture, build tooling, and developer experience infrastructure. I specialize in data-dense and real-time user interfaces, query language tooling, and large-scale application architecture, with a focus on systems that compound across teams over time. I am known for holding product judgment and technical depth at the same time, and for building things that are both structurally sound and genuinely good to use.

Skills

Platform and Architecture: React, TypeScript, Vite, TanStack Router, Webpack, single-spa, pnpm workspaces, Turborepo, module federation, feature flags

Data Visualization and Real-Time UI: React Flow (xyflow), elkjs, D3, SVG, SSE streaming, real-time state management

Editor and Query Tooling: query language UX, live parsing, autocomplete, mode detection, tokenization

Backend and Data: Node.js, GraphQL, AWS (Lambda, S3, DynamoDB), DuckDB, Neo4j, SQL

AI Engineering: Claude Code, MCP, LLM-as-judge eval, vector embeddings, AWS Bedrock, agentic workflows

Design Systems and Accessibility: MUI 5, design tokens, WCAG 2.1 AA, FedRAMP, Figma, Storybook, Vitest, Playwright

Experience

Senior Software Engineer, Applications (promoted from Software Engineer, Feb 2022)

JupiterOne, New York, NY February 2022 – Present

JupiterOne is a cybersecurity graph database platform. I joined as an early frontend engineer and have since owned the company’s application platform, design system, data visualization surfaces, and AI platform, building the shared foundation other product teams build on top of.

Platform Rebuild: 36 Micro-Frontends to Single Monorepo

Diagnosed the micro-frontend architecture’s fundamental cost in September 2024 and spent 18 months building the technical case before architecting the replacement: a single monorepo producing one Vite build artifact with new module boundaries, a shared primitive layer, and org-wide build tooling. Went live April 2026; all product teams are migrating onto it.
Designed a URL-as-source-of-truth navigation architecture using TanStack Router and outlet-composed views, including routable modals, deep-linked graph state, and inspector stacks encoded in the URL, replacing fragmented state management across 6+ teams. Shipped the inspector as a universal navigation primitive in June 2026, available from every surface in the product rather than only within query results, making every traversal shareable as a link.
Designed the module boundaries, API contracts, and migration path for the new platform, and trained all product teams through the transition, making it something teams build onto rather than around.

Graph Canvas and Data-Dense UI

Built a production-scale graph canvas as an in-house replacement for a commercial graph visualization library (est. $75k/year licensing), delivering more product-specific behavior using React Flow and elkjs at a fraction of the cost.
Designed every behavioral layer of the canvas to encode recoverable paths: entity-class packing gives users spatial memory across sessions so the same data always produces the same layout, node dimming preserves context while focusing attention on a selection, and the inspector modeled as a push-pop navigation stack so traversal never orphans the user. The canvas has no states that require a modal escape or a dead-end back button.
Introduced a directed query-path graph with a left-to-right entity-hop spine and frontier fan-out, so users can discover their data model interactively rather than requiring schema knowledge before writing a query.

AI Platform and Real-Time Engineering

Founded the company’s AI platform from a blank file in October 2025, including the SSE streaming architecture, real-time token rendering, J1QL generation agent, and the evaluation harness that all subsequent AI features run on.
Engineered a full LLM evaluation pipeline: Levenshtein-scored datasets, LLM-as-a-judge scoring, schema validation against the live query engine, Bedrock multi-region routing, and interrupt semantics, establishing the quality bar that all subsequent AI features ship against.
Built a natural language to J1QL conversion pipeline with a vector embedding cache to reduce repeat query latency and operational cost, launching with one of the highest feature adoption rates in the product’s history.
Self-initiated an AI chat feature with live page context (route watching, cache inspection, integrations context), brought it from a weekend project to a CEO demo to a shipped core product capability.
Built and documented an agentic engineering workflow using Claude Code with parallel subagents, git worktree isolation, and MCP server integration for codebase-wide tasks. Authored the internal RFC for org-wide adoption.

Query Language Editor and Search

Built a unified search input with live J1QL mode detection, including auto-classification of query vs. natural language, a correctable chip UI, schema-aware autocomplete, and live parsing, making a graph query language feel as fast as a search box.
Built a visual J1QL query builder at the schema explorer page that removes the prerequisite of schema knowledge: users navigate their account’s entity-class taxonomy as a focused subgraph, click classes to assemble traversal chains, and get the resulting J1QL generated live. The ordered path is the single source of truth; chain state, filters, and selection are URL-serialized so every partially-built query is shareable and refresh-safe.
Built SearchAnywhere across three major versions over two years, each shaped by what the previous version taught about how security engineers actually search: from a unified query entry point, to persistent history and a resizable preview drawer, to full AI integration with multiple query candidates, culminating in a GA org-wide announcement in March 2025.

Design System, Accessibility, and Developer Experience

Sole owner of JupiterOne’s production design system for three years, including token architecture, MUI 5 migration across 30+ repos, component API design with composable slots that let product teams customize behavior without forking components, versioning strategy, and org-wide developer experience.
Led the WCAG 2.1 AA compliance initiative and FedRAMP eligibility work through the design system, enabling government contract acquisition.
Built web-juno-datatable from scratch in September 2022, the org’s primary table layer, through four generations from a custom v1 to MUI DataGridPro to a lightweight variant, adapting to performance, accessibility, and product requirements at each stage without over-abstracting before the requirements were clear.
Flagged a $5,000/month bill from a hosted visual-testing vendor and replaced it org-wide with self-hosted Storybook and Vitest interaction tests in May 2026.

Full-Stack Feature Ownership

Built Dashboard Query Variables end-to-end: service design, frontend implementation, and a data migration on shipped customer data, enabling non-technical users to configure parameterized graph queries without writing J1QL.
Authored the Snapshot Series RFC, accepted org-wide in July 2026, after diagnosing that three existing platform machines had been writing full daily J1QL result rows to S3 for up to 8 years with no read layer on top of them. Designed the full pipeline: capture via existing query-service machinery, transcode to Parquet via DuckDB on EKS, rollup aggregation, and a report reader serving 32 query templates, enabling 90-day trend queries and executive dashboards on historical graph data for the first time.

Software Engineering Technical Coach

Flatiron School, New York, NY October 2021 – January 2022

Led 16 career-changers through the software engineering curriculum via pair programming, bi-weekly lectures, and project reviews.

Field Implementation Business Partner, New Stores and Remodels

Target, New York, NY June 2014 – September 2021

Directed field teams, general contractors, and trades across 9 large-scale, multi-million dollar construction projects across the US, on time and on budget, with active scope management.
Coordinated across business owners, architects, and contractors throughout the full project lifecycle.

Community

Software Engineering Mentor, Out In Tech, New York, NY (Sep 2024 – Dec 2024) Mentored 5 LGBTQ+ students in their final CS year, helping them ship and deploy real projects and network in NYC tech.

Software Engineering Mentor, Code The Dream, New York, NY (Mar 2022 – Nov 2023) Mentored 26 low-income and BIPOC students learning to code and supported them through the job search process.

Education

Utah Valley University, B.S. Digital Marketing, Magna Cum Laude Salt Lake City, UT | April 2014

Leadership Certification with High Distinction
1st Place, American Marketing Association Strategy National Competition