2022—Now
Software Engineer
Independent
Building and maintaining developer tooling, distributed desktop systems, simulation software and GPU-accelerated ML systems. Work includes Tethoq, NH Trainer, Shape Vectorizer, Klipt and glTF-to-JSON.
Selected work · 2021—2026
Kerim Yalchin
Software engineer based in London.
Commercial experience at Whitespace / Verisk and Acue, alongside independent work across TypeScript, Java, Python, applied ML and developer tooling.
01 / Experience
Professional experience
Software engineering roles at Whitespace / Verisk and Acue, followed by independent engineering work.
2022
Software Engineer · Contract
Whitespace / Verisk
Improved page loading by 10× through a memory-management change in a core algorithm, then shipped TypeScript and Angular features and resolved production issues for client-facing software.
2021—22
Software Engineer · Contract
Acue
Designed the SQL model and AWS environment for a Java/Spring sports platform, including EC2 and Cognito, while leading the resolution of delivery-blocking stories in a CI/CD team.
02 / Selected work
Selected projects
Five projects covering software engineering, ML systems, GPU work and developer tooling, ordered by depth and completeness.
01Distributed systems · Developer tooling
Tethoq
A provider-neutral workspace for local and remote coding-agent sessions.
The system spans an Electron/React desktop client, a TypeScript bridge and encrypted relay, a Flutter remote client, normalized provider adapters, browser tooling, approvals and multi-agent delegation.
- 12
- built-in integrations
- 175
- session stress fixture
- E2E
- encrypted relay
- TypeScript
- React
- Electron
- Flutter
- WebSockets
Working product · Tethoq02Flagship · Neural networks + simulation
NH Trainer
A browser combat simulator powered by custom-designed neural-network opponents.
A complete practice client built around neural-network opponents designed and trained specifically for the simulator. The project covers combat simulation, the training pipeline, policy evaluation, browser inference and the interaction system.
- 313
- policies evaluated
- 48,828
- simulated fights
- 1e-5
- score parity tolerance
- TypeScript
- React
- Three.js
- Java
- Reinforcement learning
Working product · NH Trainer03GPU computing · Graphics
Shape Vectorizer
A GPU raster-to-vector optimizer that redraws images as editable SVG.
A local tool that fits hundreds of layered vector shapes to a raster target. The fast path uses a fused CuPy/NVRTC CUDA renderer with analytic gradients, with a portable PyTorch fallback.
- 450
- editable shapes
- 11.4s
- reference run
- 25.32
- dB PSNR
- Python
- PyTorch
- CUDA
- CuPy
- SVG
Working product · Shape Vectorizer04Desktop product · Automation
Klipt
A focused desktop tool for turning long videos into precise clips.
An Electron application that coordinates yt-dlp and FFmpeg behind a small, usable interface. It supports source quality selection, exact time ranges and self-healing dependency updates across YouTube, Twitch and Kick.
- Electron
- Node.js
- FFmpeg
- yt-dlp
- HTML/CSS
Working product · Klipt05Developer tooling · 3D runtimes
glTF to JSON
A one-file converter for getting animated 3D assets into constrained game runtimes.
A Python GUI and CLI that converts glTF/GLB assets into compact, integer-quantized JSON while preserving vertex order, hierarchical skinning and stitched animation clips.
- Python
- glTF/GLB
- 3D transforms
- CLI
- Data pipelines
glTF / GLBparse + bakecompact JSON
Additional work
Client projectKing’s College Hospital CMS
Education CMS built for a King’s College Hospital team using Micronaut, Hibernate and REST APIs.
Computer visionRed Tile Player IdentifierHSV segmentation, rotated rectangles and confidence scoring in OpenCV.
Machine learningJunctional ClassifierSVM, neural network and KNN comparison on retinal vessel features.
03 / About
Background
My work spans frontend applications, backend systems, applied ML, simulation and developer tooling. I tend to work across the full path from implementation and testing to the final interface.
Software engineering
TypeScript, React, Angular, Electron, Flutter, Node.js, HTML and CSS
Backend & systems
Java, Spring Boot, WebSockets, REST APIs, SQL, relay security and data modelling
Applied ML & vision
PyTorch, CUDA/CuPy, reinforcement learning, OpenCV and OCR
Delivery
AWS EC2/Cognito, Docker, Git, CI/CD, Playwright and test automation
Education
BSc (Hons) Computer Science · 2:1
King’s College London · 2018—2021
04 / Contact
Contact
For work enquiries or questions about any of the projects, email me.
kerimyalchin@gmail.com