# Juan M. Tissera > Senior Platform & DevOps Engineer · AI Infrastructure · GCP/AWS, Kubernetes, Terraform · Edge AI at 10,000+ camera scale From multi-cloud Kubernetes to Jetson boxes at the edge. Senior platform and DevOps engineer, twenty-plus years in software, and still shipping complete products. I care about one thing: infrastructure that lets engineering teams move faster than their size suggests. Over the years that has meant a mobile studio shipping for Canon, VISA and adidas, an Intel-backed company putting AI on ten thousand cameras for a national customs agency, and today platform work for AI product teams as an independent engineer. The constant is the same. I own the architecture and production, and I work embedded with the engineers who ship on top of it: what gets built as code, what gets automated, and what is not worth doing at all are decisions we make together, then I build them, with coding agents for speed and my name on the result. I have led engineering teams as a tech director and a tech lead, and I have been the platform engineer inside product teams. Both taught me the same thing: platforms exist so other people can ship. Location: Argentina (UTC−3), remote. Languages: Spanish (native), English (advanced). Works remote, on contract. ## Experience - Platform & DevSecOps Engineer at Authentrics.ai (2025-10–present) - Wrote the full GCP estate as Terraform: 300+ resources across 9 environments (GKE Autopilot and GPU pools, Cloud Run, global load balancers, Cloud Armor, API Gateway, KMS, Secret Manager). - Brought a hand-built cloud under IaC with zero downtime: 316 resources across 9 environments in 12 modules, including 48 Cloud Build triggers, imported behind zero-diff gates. - Designed keyless multi-cloud delivery: OIDC federation into AWS for ECR push; Workload Identity Federation so partner EKS clusters pull from Artifact Registry with no stored keys. - Unified the GKE and EKS Helm charts into one: 6 services deployed to 7 environments on GKE and EKS, cut over to live pipelines with byte-identical renders. - Delivered images to 4 partner AWS accounts through Workload Identity Federation with no service-account keys, promoting the same digest from dev to prod; 25 Artifact Registry repositories hold 84 container images. - Built the CVE programme: Trivy and Grype on every image, a weekly sweep over 14 image tags across 3 environments (the first sweep surfaced 12 critical and 388 high findings) and one owned, auto-closing JIRA ticket per image. - Cut cloud spend by ~70% with cost dashboards (Flask + BigQuery), GPU scale-to-zero and idle-environment shutdown. - Lead Engineer at Neuvana AI (2026-02–present) - Built a Jetson Orin NX edge vision pipeline: YOLO11 detection, depth-aware tracking with a ZED 2i, on-device face blur, catalogue recognition. - Built a multi-tenant analytics platform (FastAPI, ~4k lines) for retail shelf and e-waste sorting, delivered as 16 work packages in ~10 days and used in enterprise presales. - Designed the "devices pull intent" control plane: no inbound execution path to edge nodes. - Created labkit, a container scaffold that generates per-node Compose files from service manifests with memory budgeting and gives each app its own private mesh network via a sidecar; 3 reference services; used it to cut a 3-service demo platform over with zero route changes. - Deployed the containerized vision pipeline across a 5-node edge fleet (3 Jetson Orin NX, an x86 GPU gateway, a CPU kiosk): ~8–9 fps depth-aware detection on Jetson at 720p and ~10 fps CPU-only on a NUC. One live camera pipeline; the other nodes ran on synthetic frames. - Solved and documented four Jetson/L4T container issues (pip index, iptables raw, cuBLAS shadowing, buildx). - Wrote the privacy architecture and DPIA package, with the guarantees enforced in code. - Built the trilingual company site and ops panel (Astro 5 SSR, SQLite/Drizzle, 38 tables): Gemini content pipeline publishing through the GitHub Trees API, lead-gen automation, deal calculator. - Talent & Technical Assessment Advisor at Mission.dev (2021–present) - Design and review technical assessments: coding challenges, multiple-choice questions and role-specific evaluation criteria. - Review candidate code-challenge submissions for solution quality, technical proficiency, problem-solving approach and adherence to requirements. - Advise talent-acquisition teams on candidate evaluation and technical screening decisions. - Develop assessment frameworks tailored to technologies, seniority levels and client requirements. - Identify assessment-integrity concerns, including inappropriate use of external assistance or AI-generated solutions, and support consistent review standards. - Help improve vetting processes and evaluation methods so assessments stay high-quality, fair and scalable. - Tech Lead, Edge AI Infrastructure at Austin GIS (2022–present) - Led the hyperconverged edge platform (compute, storage and inference on one cluster) behind a 10,000+ camera customs deployment across 198 sites in Mexico: 90-day retention at 1080p/15fps on MinIO with erasure coding, encryption and site replication. - Engineered the far-edge node stack: Intel NUC, Ubuntu 22.04, K3s, containerized video and audio analytics, vPro AMT and MeshCentral remote management, automated onboarding. - Reached 300+ VMS streams per server, 60+ with AI analytics, with RAFT-based clustered failover. - Delivered the white-label AustinGIS VMS on Axxon One plus its integration SDK (HTTP, gRPC, WebSocket events). - Deployed smart-city analytics across three Mexico City municipalities (6,252 cameras) and a Lima mall retail pilot (200+ endpoints, 90% reliable demographic segmentation). - Built OpenVINO, DL Streamer and Intel Geti / SceneScape pipelines: weapon detection, man-down, loitering, people counting, face recognition, person re-identification. - Digital-signage spin-off (2026–): built fleet management for 400 Android signage tablets: bulk policies, device traceability and issue views. - Digital-signage spin-off (2026–): designed a vendor-neutral device-management proxy that kept the fleet operable through a 12-hour vendor API outage. - Digital-signage spin-off (2026–): isolated 3 environments (production, UAT, depot) on one shared device registry, with deployments that verify the live outcome after each release. - Senior Full-Stack Engineer at Intel (2020–2021) - Built IoT and video-AI software for smart-city programmes: model training for video analysis, custom OpenVINO and DL Streamer plugins. - Tuned Linux for automated edge-client onboarding; fleet management with vPro AMT, MeshCentral and MeshCommander. - Tech Director at MobyDigital (2013–2020) - Managed 5 teams of about 4 engineers each (~20 people). - Led architecture, initial coding and deployment of mobile apps for Canon Europe, VISA, ICBC, the Dutch government and retail brands. - Tech Director · Software Developer · Web Developer at Brandigital (2003–2013) - Tech Director (2009–2013): bespoke CMS and CRM products rolled out across 10 countries for adidas, Mastercard, VISA, Warsteiner, Club Med, McDonald's and Coca-Cola; analytics dashboards, bulk email, notifications. Python, Node.js, AWS, PostgreSQL, MySQL, Redis. - Software Developer (2005–2009): built the core marketing product, 10M+ users; World Cup campaign widgets. AWS, PHP, .NET, JavaScript. - Web Developer (2003–2005): promotional sites and news portals. .NET, JavaScript, MySQL. - Earlier - Full-stack Developer, Anden Solutions LLC (2007–2009): EdTech collaborative forum tied to NASA STEM resources. PHP, Joomla, MySQL. - Java Developer, Griensu (2002–2003): ERP development. Java, JSP, Struts, Hibernate, WebSphere, PostgreSQL. - Head of Systems (pro bono) at Fundación Goodwill (2022-03–present) - Designed and built GWAPP, the platform behind the trips: intake, clinical records with SOAP episodes, prescriptions, pharmacy dispensing, equipment loans and travel logistics. - Ran a self-hosted Supabase platform (13 services on Coolify) serving four apps: 68 tables, 93 migrations, 177 row-level security policies. - Integrated the apps through an append-only domain-event bus plus Realtime, so field apps work offline on SQLite and sync later. - Added AI where it saves manual work: Gemini RAG assistant on pgvector, automatic clinical summaries, visit transcription, historical PDF extraction. - Built the legacy-archive migration (PDF and XLSX parsers, duplicate detection, assisted merge) and the public site (Astro SSR, anti-bot forms). ## Skills - Cloud: GCP (GKE, Cloud Run, Cloud Build, Cloud Deploy, Artifact Registry, KMS, Secret Manager, Cloud Armor, API Gateway, BigQuery), AWS (EKS, ECR, IAM, OIDC) - Platform: Terraform, Kubernetes, K3s, Helm, skaffold, Argo CD, Docker, Docker Compose, Linux, systemd - CI/CD: Cloud Build, GitHub Actions, Cloud Deploy, Dokploy, Coolify - Security: DevSecOps, Trivy, Grype, CVE management, Workload Identity Federation, OIDC, RBAC, RLS, GDPR/DPIA - Networking: Traefik, nginx, WireGuard, Tailscale, load balancing - Data: MongoDB Atlas, PostgreSQL, Supabase, pgvector, SQLite, MySQL, Redis, MinIO (erasure coding, site replication) - AI / CV: NVIDIA Jetson, CUDA, YOLO11, PyTorch, ONNX Runtime, OpenCV, OpenVINO, DL Streamer, GStreamer, Intel Geti, RAG, Gemini API - Languages: Python, TypeScript, JavaScript/Node.js, Bash, HCL, SQL, C++ - Frontend: React, Astro, Tailwind CSS - Video / IoT: VMS (Axxon One), MQTT, vPro AMT, MeshCentral, Android Enterprise MDM ## Links - Site: https://bitof.dev - Resume (JSON Resume): https://bitof.dev/resume.json - GitHub: https://github.com/juanzete - LinkedIn: https://www.linkedin.com/in/jmtissera Source of truth: docs/PROFILE.md, exported 2026-10-08T16:42:46+00:00.