Yegor Gaidar
Available · Q4 2026
Résumé · updated 22 October 2026

Yegor Gaidar
Distributed systems
& AI infrastructure.

Eight years designing and operating production systems — real-time pipelines, container orchestration, zero-trust networking, self-hosted AI. Paris, Île-de-France. Open to senior and staff engineering roles, hybrid or remote.

Résumé (PDF)

This page prints as it stands · A4 · Ctrl / ⌘ + P

8 years
Production systems
12 nodes
Self-built Nomad cluster
932K+
Sensor points in production
0.6%
CodinGame, worldwide
Profile

Engineer with eight years across the stack — from real-time embedded firmware to self-hosted, cloud-native infrastructure. Streaming pipelines (Kafka, ClickHouse), scheduling and orchestration (Nomad, Kubernetes), service discovery and mutual TLS (Consul, ACLs, Workload Identity), infrastructure as code (Ansible, Proxmox), observability (Prometheus, Grafana) and applied AI (PyTorch, on-device and self-hosted LLMs).

Advanced Master in Artificial Intelligence, Télécom Paris. The work that suits me best is a system with real constraints — a deadline, a data boundary, a bill — where the answer has to be measured rather than asserted.

Contact
Location
Paris, Île-de-France
Availability
immediate
Email
maingaidar [at] gmail.com
Web
yegorgaidar.org
GitHub
gaidar0yegor
LinkedIn
yegor-gaidar
Work mode
hybrid · on-site · remote
Experience

Six roles, in
reading order.

One column, most recent first, dates in the margin. The PDF beside this page is generated from the same text, so an application system parses this document and a human reads the same words.

Mar 2026 — present
NetIntel · Île-de-France · remote

Lead Developer & Founder

DePIN sensor-data platform

  • Real-time distributed pipeline ingesting, normalising and geolocating 932K+ multi-sensor points (WiFi, cellular, BLE, IoT) across France.
  • Go/Gin API (30+ endpoints) → Kafka KRaft → ClickHouse (MergeTree, materialised views) · PostgreSQL · H3 geo-indexing at resolutions 5–9 · Flutter client.
  • Multi-layer anomaly and fraud detection — velocity, frequency, plausibility, signal — with Isolation Forest on top.
  • Server-side vector tiles, GPU-accelerated clustering and a 3D globe projection for the map.
  • Cloud Run, GitHub Actions CI/CD, Sentry for Go and Flutter, 253 passing tests.
Dec 2025 — Mar 2026
Shneider’s Villemomble · Villemomble

Head of Digital

Four-month contract

  • Automated the ERP / CRM / WMS chain with n8n, removing the recurring manual re-entry of the same records across all three systems.
  • Deployed an internal self-hosted AI stack — OpenWebUI, open-weight models, a vector database — keeping company data on company infrastructure.
  • Integrations: WhatsApp Business API and MCP-driven task orchestration.
Oct 2024 — Aug 2025
Planète Technologie · Paris

Technical Lead

Eleven-month mission

  • Re-architected and scaled a product-aggregation platform to 2M+ references, with multi-source ingestion and a multilingual AI content-generation pipeline.
  • Owned architecture decisions, ran code review and mentored the development team.
Oct 2023 — Jun 2024
RDI Manager · Trappes

Full-Stack Engineer / AI Lead

Nine-month contract

  • Built an AI platform for research-notebook analysis — NLP, LLMs, vector search, dashboards — as microservices with a semantic indexing pipeline.
  • Replaced keyword lookup over research notebooks with semantic retrieval, so researchers find prior work by meaning rather than exact wording.
Sep 2020 — Jan 2023
SAYS Logistics · Compiègne

Full-Stack Web Developer

  • Real-time logistics tracking: IoT sensors, automated alerting, operational dashboards.
  • Automated the data flow between IoT ingestion and the dashboards operations actually used.
Sep 2018 — Aug 2020
L’Âge Bleu · Mouy

Mobile Developer

  • Mobile applications for retirement homes: Stripe payments, SMS notifications, accessibility-first interfaces for elderly users.
Selected infrastructure

The cluster I run
as the proof.

Not a client engagement: personal infrastructure, operated continuously, used as the reference for the reliability work on fractional SRE.

2026 — running
Bare metal + Proxmox · two sites

Nomad + Consul cluster

  • Three-node control plane on Raft with a 12-worker fleet — 76 threads, 124 GB RAM — provisioned end to end with Ansible.
  • Assembled from nine second-hand Mac minis and a used dual-Xeon server: no monthly cloud cost.
  • Zero-trust throughout — gossip encryption, RPC and HTTP mutual TLS, ACLs, Consul Workload Identity issuing per-service tokens.
  • Prometheus and Grafana across 21 targets with alerting to chat; self-hosted GitHub Actions runners.
  • Workloads reschedule after a node failure in ~15 s; bare-metal onboarding is automated with a Debian preseed.
2026 — public
llama.cpp / vLLM · one 8 GB GPU

Inference benchmark

  • Measured a single 8 GB card serving a 3B model: +87% aggregate throughput and a 20× shorter p95 time-to-first-token from serving configuration alone.
  • Published with raw results and the method, including where a hand-written kernel was not worth the effort — the write-up.
Skills

Tools by the job
they do.

Grouped by problem, not by fashion. Ordered by depth of use.

OrchestrationNomad, Consul, Ansible, Kubernetes/k3s, Docker, Proxmox
Data & streamingKafka (KRaft), ClickHouse (MergeTree, materialised views), PostgreSQL 16, Redis, MongoDB, GraphQL
Security & networkMutual TLS, ACLs, gossip encryption, Workload Identity, ansible-vault, JWT, OWASP, Tailscale, WireGuard
Observability & CIPrometheus, Grafana, GitHub Actions (self-hosted runners), Sentry
ReliabilitySelf-healing scheduling, health-checked service discovery, automatic failover and reschedule, fault injection
LanguagesGo, Rust, Python, C++, TypeScript, JavaScript, Dart, C#, Bash, SQL
AI & MLPyTorch, TensorFlow, LLM serving (llama.cpp, vLLM, Ollama), QLoRA fine-tuning, NLP, vector search, MLOps, Isolation Forest, DBSCAN, n8n, MCP
Embedded & IoTESP32/ESP8266, STM32 / ARM Cortex, I2C/SPI/UART, IMU, BLE, RTOS, real-time firmware
Backend & APIGo (Gin, pgx/v5, gorilla/websocket), FastAPI, Django, NestJS, OpenAPI 3.1
Frontend & mobileReact, Next.js, Three.js, WebGL, WebAssembly, Flutter 3.x, Riverpod
Geospatial & Web3H3 (Uber), vector tiles MVT/protobuf, trilateration, haversine, Mapbox · Solana (Anchor/Rust, Token-2022), PDAs, DePIN tokenomics
Education
Télécom Paris
Advanced Master in AI · Data & MLOps · 2022–2023
Institut F2I, Vincennes
Bachelor, web & mobile development · 2020–2022
Promeo / UIMM
BTS SIO, option SLAM · 2018–2020
The Advanced Master was full-time study, in the middle of the timeline above rather than after it.
Languages & additional
French
fluent
English
professional, B2+
Ukrainian
native
Russian
native
CodinGame
top 0.6% worldwide
Availability
immediate
References available on request. Engagement terms — scope, milestones, ownership — are on /hire.html.
This page is the screen version of the same document: one column, no floats, dates in the margin, so a parser and a recruiter read the same order. The PDF at /resume.pdf is print-ready A4.
Contact

Open to the right
full-time role.

Senior or staff engineer, systems and infrastructure, Paris or remote. Engagement terms are below, and the PDF is this document in print.

Download the PDF maingaidar [at] gmail.com
ElsewhereGitHub
NetworkLinkedIn
ProcessHiring
Based inParis, Île-de-France
Yegor Gaidar · Privacy · 2026