Tanguy PEMEJA

Senior Backend & Platform Engineer

Tokyo → Singapore, early 2027

Summary

Backend and platform engineer with 5 years building Python and FastAPI APIs and AWS-based infrastructure for edge and cloud systems in IoT and cybersecurity. Sole maintainer of the device-management API running on an edge AI fleet, and designer of its pluggable remote-access layer. Previously cut new-customer deployment from 3-4 days to 2 hours at an XDR security vendor, and led APAC incident triage after being selected to open its Tokyo office.

Experience

Senior Backend Developer — EDGEMATRIX Inc

Case study →

Tokyo, Japan · May 2024 – Present

Edge AI platform: on-site Linux stations run AI video-analytics apps from a cloud marketplace.

  • Own the FastAPI device-management API on each edge station as sole maintainer, having built most of its operations: networking, time, users and API tokens with RBAC, and licensing.
  • Designed pluggable local and remote access (Cloudflare Tunnel, ngrok support sessions, custom DNS) with shared retry, reboot recovery and runtime Traefik/Authelia routing, letting support maintain stations remotely instead of on site.
  • Extended the Python D-Bus host service (Chrony, stack upgrades, license sync) and took live upgrade progress over WebSocket from proof of concept to production; maintain the station’s Docker Compose stack (Traefik, Authelia) and CI.
  • Maintained cloud services on Elastic Beanstalk for device lifecycle, configuration and synchronization across the station fleet, with a staging environment and CloudWatch alarms; contributed to the encrypted, device-specific update pipeline (S3 presigned URLs, short-lived ECR credentials).
  • Migrated internal cloud platform login from GitHub to Auth0 (OAuth2/OIDC), adding Google sign-in for staff without GitHub accounts and finer-grained roles.
  • Extended on-station observability (Grafana, Loki): container-health and AI pipeline dashboards (e.g. processed video FPS), new log sources and log export for support; prototyping a read-only MCP server for LLM-assisted diagnosis.

Cybersecurity Engineer — TEHTRIS

Case study →

Tokyo, Japan · Dec 2022 – May 2024

European XDR cybersecurity vendor (EDR, EPP, SIEM, SOAR).

  • Co-built, with an engineer in France, a self-service platform for the deployment team that provisions a full multi-server customer instance on OVH Cloud (Airflow, Packer), cutting deployment from 3-4 days to about 2 hours.
  • One of six people selected to open the Tokyo office: led incident triage during APAC hours for the three-person team, deciding when to escalate production issues and regional customer escalations to engineering in France or Canada, or directly to the CTO; extended Grafana dashboards and alert thresholds.

Junior R&D Cybersecurity Consultant — TEHTRIS

Case study →

Pessac, France · Sep 2021 – Dec 2022

Same company, before the internal transfer to the Tokyo office.

  • Built a FastAPI layer exposing the XDR platform’s internal APIs to customers and external tools: reviewed the API contracts, simplified and secured the calls, generated the docs from the OpenAPI spec and covered them with Postman test suites.
  • Built, as sole developer, an internal operations platform (FastAPI, HTML/CSS) giving support teams and staff abroad scoped access to OVH Cloud environments through Okta login and custom roles and actions.

Machine Learning Intern — INRIA

Bordeaux, France · Mar 2021 – Aug 2021

French national research institute for digital science and technology.

  • Neural-network control models for a prosthetic arm, with a VR simulation environment for data collection and evaluation.

Selected projects

  • FastSpec (fastspec.kaseovo.com): Open-source app (AGPL-3.0) for writing, versioning, diffing and linting OpenAPI specs, with an 11-tool MCP server for AI agents; ships as one Docker image or serverless on AWS (CDK).
  • Instagram Agent (demo: dashboard-demo.kaseovo.com/insights): Serverless AWS backend (Lambda, SQS, DynamoDB, Terraform) that ingests Meta webhooks and uses Gemini to tag comments and draft replies, with a human confirming every action.
  • Belote AI (github.com/tpemeja/belote-ai): Found and fixed a seeding bug in Google DeepMind’s OpenSpiel (merged upstream) while building card-game agents; a neural agent matches Monte Carlo search’s play at 0.9 ms instead of 120 ms per deal.

Technical skills

  • BackendPython, FastAPI, Pydantic, Flask, REST, WebSockets, OpenAPI
  • Cloud & InfrastructureAWS (Lambda, Elastic Beanstalk, S3, ECR, SQS, DynamoDB, CloudWatch), OVH Cloud, Docker, Docker Compose, Terraform, AWS CDK, Packer, Airflow, GitHub Actions
  • Edge & NetworkingLinux, D-Bus, Traefik, Cloudflare Tunnel, ngrok
  • Observability & OperationsGrafana, Loki, Prometheus, PagerDuty, incident response
  • Identity & SecurityOAuth2/OIDC, RBAC, Auth0, Okta, Authelia, XDR (EDR, SIEM, SOAR)
  • DataPostgreSQL, MySQL, Redis, DynamoDB, SQLite
  • AIMCP, Gemini, PyTorch
  • Frontend & TestingVue.js, PrimeVue, pytest, Postman

Education & certification

  • Diplôme d'Ingénieur (MSc Computer Science)ENSEIRB-MATMECA · 2018–2021
  • Exchange semesterKeio University · 2020–2021
  • AWS Certified Solutions Architect – AssociateAmazon Web Services · Verify

Languages

  • FrenchNative
  • EnglishFull professional
  • JapaneseElementary