I'm a full-stack engineer and technical lead who enjoys turning customer problems into working products. My work spans React/Next.js interfaces, TypeScript and Python services, APIs, data pipelines, and cloud deployments.
I'm looking for a full-time product, full-stack, founding-engineer, or backend role on a small team, bringing my founding-engineer experience at Polyhistor to a new product. Based in Washington, DC; open to remote work and relocation, including SF and NYC.
My hands-on work spans the TypeScript/React pilot interfaces, Next.js website, backend services, geospatial database architecture, ingestion pipeline, and APIs.
- Shipped neighborhood-insight pages and walkability, greenery, and noise-scoring APIs for Inhotel Inc.'s European-city pilot.
- Built custom Mapbox modules and D3 visualizations for location insights.
- Built a DuckDB pipeline over 3.55M+ Overture Maps places, avoiding the memory failures of the earlier Pandas approach.
- Combined PostgreSQL/PostGIS, pgvector, Sentence-Transformers, and HNSW in D-GIX for spatial and semantic search; added Redis embedding caching, persistent HTTP connections, and batch lookups.
D-GIX benchmark results and conditions
| Result | Evaluation context |
|---|---|
| 0.97 Recall@10 | 70 curated semantic queries; category-based relevance |
| 19ms median warm-cache search | Repeat queries with embeddings cached in Redis |
| 116ms median first-time search | Query embedding uncached; model already loaded |
| 4,414 requests, zero failures | Separate 100-user, 60-second test on a read-heavy cached endpoint |
| 98.7% lower modeled cost vs. Google Places | Pricing scenario of one million monthly requests |
These are benchmark and modeled-cost results. The cached-endpoint test does not establish zero failures across the entire API. The current density signal is inferred from place attributes.
|
An independent, one-book retrieval study comparing BM25, dense retrieval, DSI, and structure-aware generative retrieval. Frozen metrics, an executed notebook, reproduction scripts, and documented limitations. Python · Retrieval evaluation · Reproducibility |
🛠️ ShipSolidA code-audit CLI that combines Semgrep findings, Claude analysis, and PDF reporting. Turns technical scan findings into reports that founders and engineering teams can review. TypeScript · Node.js · Applied AI |
|
A browser-based LaTeX editor with a PDF compilation service and heuristic resume/JD analysis. A React/TypeScript interface, Express API, and local Docker setup. React · TypeScript · Node.js · Docker |
The Next.js website for Polyhistor's location-intelligence platform. Connects the platform's technical capabilities to a public product experience. Next.js · TypeScript · Product delivery |
Also available: the earlier Polyhistor consumer-app backend, a public fork using Node.js, Express, TypeScript, PostgreSQL/PostGIS, JWT, and Socket.IO. It is separate from the D-GIX search architecture.
- Frontend & visualization: React, Next.js, React Native, TypeScript, JavaScript, HTML5, CSS, Mapbox, D3
- Backend & APIs: Node.js, Express, Python, Django, Java, Spring Boot, REST, GraphQL, WebSockets
- Data & search: PostgreSQL, PostGIS, pgvector, MongoDB, Redis, DuckDB, H3, embeddings, HNSW
- Cloud & quality: AWS, Docker, Kubernetes, Terraform, GitHub Actions, Jest, Playwright, Storybook
- AI & developer tools: LLM integration, RAG, Sentence-Transformers, Cursor, Claude Code, Codex
Under the hood: the Polyhistor search stack
The search system combines semantic retrieval with spatial filtering. Redis caches query embeddings, while PostgreSQL brings relational, spatial, and vector data together.
flowchart LR
Client["Web interfaces / API clients"] --> API["Node.js / TypeScript API"]
API --> Cache["Redis embedding cache"]
API --> Embeddings["Python embedding service"]
API --> DB["PostgreSQL + PostGIS + pgvector"]
Source["Overture Maps places"] --> Ingest["DuckDB ingestion"]
Ingest --> DB
style API fill:#143b45,stroke:#2dd4bf,color:#ffffff
style DB fill:#172c46,stroke:#38bdf8,color:#ffffff
style Embeddings fill:#272f4d,stroke:#a78bfa,color:#ffffff
Experience & education
Polyhistor Inc. — Software Engineer & Technical Lead
Jan 2026–Present
Lead platform architecture and hands-on full-stack delivery with a small engineering team; helped lead the pivot from a consumer travel product to an enterprise geospatial API and customer pilot.
Rebecca Everlene — Software Development Engineer II
Jan 2025–Jan 2026
Led Landmarks from concept to public beta in four months while coordinating 12 engineers across React Native and Node.js. Built a Storybook component library that cut feature delivery time by 30%, and established CI/CD and review standards.
Grid Dynamics — Software Engineer
Dec 2022–Sep 2023
Built Java/Spring Boot services processing 50K+ daily transactions at 99.9% uptime and a GraphQL layer that improved screen load times by 30%. Led a 150K+ line TypeScript migration and reduced Android crash rates by 15%.
IG Defence — Software Engineer
Jul 2020–Dec 2022
Rebuilt OAuth/JWT authentication and dashboard flows across 10,000+ business accounts, increasing retention by 50%. Built WebSocket/Redis notifications that reduced alert latency to under two seconds.
M.S. in Computer Science, Illinois Institute of Technology — May 2025
Bachelor's degree in Computer Science, Bharath Institute of Higher Education — 2023
Let's build something useful.
LinkedIn · Email · Polyhistor