DA

Full-Stack Software Engineer · Applied AI Engineer

I build full-stack software and the AI that powers it.

I’m David Aihe. I spent more than three and a half years building enterprise applications at JPMorgan Chase, from Spring Boot services to React interfaces. Now I ship agentic and retrieval systems end to end.

View my workRésumé
Portrait of David Aihe, smiling

Open to work

3.5+ yrs at JPMorgan Chase

Selected work

Things I designed, built, and shipped.

Six projects, from desktop tools to multi-agent systems. Every screenshot is the real app.

Jane settings window on the Dictation tab, showing the push-to-talk key and hold timing

Windows desktop app

Jane 1.0

Hold a key, speak, release. The text lands wherever you were typing.

System-wide push-to-talk dictation that runs entirely on your own machine. Speech recognition, clean-up, and voice editing all happen locally, so no audio ever leaves the device.

  • Recognizes a one-second utterance in 80 ms, about 18× faster than real time, on the CPU
  • Edit Mode rewrites selected text from a spoken instruction
  • Reads the focused window so names and jargon come out spelled correctly
  • C#
  • .NET 10
  • WPF
  • Parakeet-TDT
  • Ollama
  • SQLite
DownloadSource
Taskyard first-run card on frosted glass over the desktop wallpaperTaskyard tools widget showing the Tasks tab with three to-do itemsTaskyard tools widget showing the focus timer in the dark theme

Windows desktop app

Taskyard

A desktop organizer for Windows 11 with glass groups and a focus timer.

Taskyard sits on the desktop layer and sorts your icons into named glass groups you can move, resize, roll up, and hide. A floating widget holds a to-do list, countdown timer, and stopwatch.

  • Seats one window per monitor directly above the Windows desktop through Win32 calls
  • Tracks files by NTFS file id, so renaming in Explorer keeps every icon in place
  • Real blur over its own copy of the wallpaper holds 60 fps while dragging
  • Electron
  • React
  • TypeScript
  • Win32
  • koffi
DownloadSource
CollabBoard canvas with an AI-generated SWOT analysis and sprint retrospective

Web app

CollabBoard

A real-time collaborative whiteboard with an AI agent that builds the board for you.

An infinite canvas with sticky notes, shapes, frames, and connectors, synced live between everyone on the board. Ask the AI assistant for a SWOT analysis or a retrospective and it lays one out.

  • Multiplayer cursors and presence with real-time state sync
  • Planner and executor agent turns plain language into board objects
  • Idempotent requests and server reconciliation prevent duplicate objects
  • React
  • TypeScript
  • Konva
  • Firebase
  • Vercel
Live demoSource
Nerdy Ads campaign dashboard with the accepted ads panel and the generate ads controls

Multi-agent system

Nerdy Ads

Four AI agents that write, score, and rewrite ads until they pass.

An autonomous ad generation engine for SAT prep campaigns. A researcher, writer, evaluator, and editor work through an orchestrator, and only ads that clear the quality bar survive.

  • Accepts an ad only at a 7.5+ weighted score with every dimension above 6.0
  • Evaluator calibrated against a tiered set of real, long-running ads
  • Covered by 240 unit, integration, and evaluation tests
  • TypeScript
  • Gemini
  • React
  • Vite
  • Langfuse
  • Railway
Source
FortranLens search screen inviting a question about the gfortran codebase

RAG application

FortranLens

Ask plain-English questions about 250,000 lines of compiler source.

Code intelligence for the gfortran compiler. It retrieves the relevant source, then answers with explanations that cite the exact file and line they came from.

  • RAG Fusion expands each question into variants and blends the rankings
  • Every answer is grounded with file and line references
  • Maps dependencies, callers, and the impact of changing a function
  • Next.js
  • TypeScript
  • Pinecone
  • OpenAI
  • Tailwind CSS
Source
Pocket Meadery batch screen for Wildflower Traditional with reminders, ABV figures, and timelinePocket Meadery dashboard listing upcoming reminders and active batches

iOS app

Pocket Meadery

A local-first brewing log for tracking mead from first pitch to bottle.

Track batches, log each step with gravity readings, and get reminders when something needs attention. Everything is stored on the phone in a relational schema designed to sync later.

  • Normalized SQLite schema with six tables and hand-written migrations
  • Goal, expected, and current ABV calculated from gravity readings
  • Local reminders scheduled from preset brewing templates
  • React Native
  • Expo
  • TypeScript
  • SQLite
Source

About

Enterprise software for 30,000+ users.

I spent more than three and a half years at JPMorgan Chase building and modernizing enterprise applications across the stack: Java and Spring Boot services, React and TypeScript interfaces, AWS deployments, and the tests that let teams ship with confidence.

In 2026 I completed the Gauntlet AI engineering fellowship and have been building applied AI products since: multi-agent pipelines, retrieval over legacy code, voice-first desktop tools, and the evaluations that prove they work.

I care about systems that hold up under load, code that the next engineer can read, and numbers that back up the claim.

building enterprise software at JPMorgan Chase
3.5+ yrs
users on the applications I supported
30,000+
faster app performance after cutting redundant calls
30%
test coverage on the features I owned
90%+

Skills

The full stack, with AI built in.

The tools I reach for, grouped by the job they do.

Certification

  • AWS Certified Developer – Associate

Experience

Enterprise scale, then applied AI.

  1. Sep 2026 – Present

    Software Engineer

    Sandstorm DesignContract

    • Delivering client web applications and CMS-driven sites on Next.js, taking changes from requirements through to production release.
    • Building reusable Sitefinity widgets and React components that connect CMS content and configuration to the application.
    • Supporting CMS platform modernization and upgrades across multiple client sites, adapting implementations and validating each change through release.

  2. Feb 2026 – Apr 2026

    AI Engineering Fellowship

    Gauntlet AITraining

    • Intensive applied AI engineering fellowship focused on RAG, knowledge graphs, AI agents, evaluations, cloud deployment, and enterprise AI systems.
    • Built and deployed ClawdForge, an agentic software factory using LangGraph and the Claude Agent SDK.
    • Built and deployed FortranLens, a RAG platform over 250,000+ lines of legacy Fortran using Pinecone and OpenAI embeddings.
    • Shipped multimodal and scheduling applications using Claude, OpenAI, PostgreSQL, and Docker.

  3. Oct 2022 – Jun 2026

    Full Stack Software Engineer

    JPMorgan ChaseFull-time

    • Built Java and Spring Boot REST APIs for internal applications, improving data reliability and supporting the teams that depended on those services.
    • Built reusable React and TypeScript components with Redux, reducing duplicate web calls and improving application performance by 30%.
    • Used BlazeMeter to test performance and support changes that improved throughput and reliability by 20% for more than 30,000 users.
    • Added Cypress and JUnit tests across assigned features, maintaining 90–100% test coverage where measured.
    • Helped modernize a company-wide internal platform by replacing legacy patterns with TypeScript and shared components.
    • Led React training sessions and worked with product owners, designers, and engineers through planning, code reviews, and delivery.
    • Proposed and demonstrated AI-assisted development workflows, including a knowledge-graph approach for tracking context across a large application.

  4. Oct 2021 – May 2022

    Full Stack Web Development

    BloomTechTraining

    • Intensive full-stack web development program covering HTML, CSS, JavaScript, React, Node.js, Express.js, and SQL.

Get in touch