live · runs in this browser · 0 servers · model loading
live · runs in this browser · 0 servers · model loading
Mehmet Emin Gench — Forward Deployed Engineer — AI Agents.
AI / Software Engineer with 8+ years of development experience and a B.Sc. in Aerospace Engineering. Builds production-grade systems across AI agents, computer vision, backend, and edge—from RAG/knowledge-graph retrieval and multi-agent orchestration to hardware-to-cloud integrations. Currently deploys enterprise AI agents for Fortune 500 companies, navigating diverse tech stacks and delivering measurable business outcomes.
Remote (Vancouver, BC).
B.Sc. Aerospace Engineering — Turkish Air Force Academy, Istanbul, Turkey.
Former Air Defense Officer in the Turkish military.
Forward Deployed Engineer — AI Agents · March 2026 – Present · Remote (Vancouver, BC)
Leading generative AI platform for customer experience. Real-time AI coaching for frontline teams, autonomous AI Voice & Chat Agents, conversation intelligence across 30+ languages. Customers include Fortune 500 companies such as Intuit, Hilton and Alaska Airlines, and Cox. Founded by Stanford AI researchers, backed by Sequoia, a16z, Greylock.
Build and deploy production-grade AI agents for enterprise customers, writing bespoke integrations and custom tooling tailored to diverse tech stacks and business requirements.
Own end-to-end delivery—from scoping and architecture through prompt engineering, rigorous model evaluation, and production monitoring—driving measurable business outcomes.
Design multi-step agentic workflows with guardrails, knowledge pipelines, and observability, bridging customer needs with core platform capabilities.
Sep 2023 – Mar 2026 · AI Engineer, Goodfintech Ltd. — Designed a multi-tenant AI Agent architecture with strict user isolation, enabling reliable execution across microservices.
Jun 2022 – Jun 2023 · Machine Learning Engineer, Vivoo — Re-engineered image processing workflows into AWS Step Functions, leveraging parallelism to boost throughput by 300%.
Oct 2020 – Jun 2022 · Full-stack AI Engineer, Novit AI (aka Archangel Autonomy) — Developed a full-stack satellite imagery platform using Django and Vue.js to detect and track maritime vessels.
Jan 2020 – Oct 2020 · Python Developer / Software QA, E-Kalite Software (PharmaCircle) — Automated data validation using Selenium, increasing QA efficiency by 70%.
2016 – 2020 · Founder & Lead Developer, Indie Game Developer (emdi_apps) — Developed and published 10+ hyper-casual games using Unity and C#, reaching thousands of organic users and managing full product lifecycle from design to store deployment.
even_glasses — G1 BLE SDK (Python) · Python · 78 stars
telegramGPT — AI bot building guide · Python · 52 stars
G1 Voice AI — Voice assistant · Python · 25 stars
g1_flutter — Flutter BLE bridge · Dart · 18 stars
visionlink — Multi-device OS · C · 11 stars
even_glasses_redis_control — Jupyter Notebook · Jupyter Notebook · 7 stars
llm_adaptive_router — LLM routing · Python · 6 stars
smart_glass_mcp — AI agent connector · TypeScript · 4 stars
AI/ML — LLMs, Agents, RAG
Languages — Python, TypeScript, Dart
Backend — FastAPI, Next.js, PostgreSQL
Infra — Docker, AWS, Linux, CI/CD
Aug 15, 2026 · The gauntlet loop — stop letting your AI grade its own homework — A post-deliverable QA ritual: fan out independent critics, compare against real examples, and recurse until zero open critiques.
Jul 12, 2026 · Building AI agents that ship — lessons from the field — Lessons learned building and deploying AI systems across four companies — from guardrails and monitoring to shipping fast.
GitHub · github.com/emingenc
LinkedIn · linkedin.com/in/emingench
X · x.com/emingench
Open to open-source and smart-glasses projects. Not taking new roles.
Loop · each question runs 4 steps. plan: pick the tool; act: run it on his notes; observe: read what it returned; evaluate: check it answers the question.
Router · Needle (26M parameters) picks the tool. A 36 MB download, run in this tab.
Chat model · SmolLM2 (360M parameters) is optional: it phrases answers. A 278 MB download (369 MB without WebGPU), loaded only if you allow it.
Check · a model draft that his notes do not back is dropped.
Privacy · both models run in your browser. 0 servers, 0 API calls, no analytics on the home page.
Speed · reading a notes file takes about 1 ms (measured 2026-10-17 in Chromium on a laptop, 6 questions).
Tests · more than 500 automated tests cover the router, loop, answers and pages.
Source · the built site is public at github.com/emingenc/emingenc.github.io. The source repository is private.