Skip to content
View giladresisi's full-sized avatar

Block or report giladresisi

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
giladresisi/README.md

Hi πŸ‘‹, I'm Gilad Resisi

Senior AI-First Fullstack Engineer and Builder from Israel

Github LinkedIn Substack


  • ⚑ My experience: 16yrs of building production SW, both large-scale (as SDE @ AWS) and SaaS (as founder @ Payroll-AI)
  • βš’οΈ I'm currently working on: Scaling Postiz, the open-source social media scheduler (28+ channels), as a core engineer
  • πŸ“– I'm currently learning: Securing LLM agents against prompt injection β€” see agent-approval-gate, a deterministic approval gate with a measured 0/6 breach rate vs 3/6 unprotected

Tech stack

My Skills


About Me

🀝 Why work with me?

- πŸ™‹ I take full end-to-end ownership on my tasks & responsibilities, beyond expectations

- ▢️ I'm biased for action, never delaying 2-way-door decisions, updating on the fly

- πŸ’¬ I communicate, directly and by async documentation. Everything must always be clear

- ❓ I always start with why, both when assigned and assigning a task, all must be justified

- πŸ‘¨β€πŸŽ“ I'm a quick & independent learner. I dive deep and reach quick results in unfamiliar topics

- πŸ‘₯ I'm a team player. I take collaboration, empathy & vibe as key for long-term success

- 🀩 I'm passionate about SW, AI, problem solving, learning, customer satisfaction, challenges

- πŸ“ˆ I love systems and I constantly make efforts to improve the systems I work with

πŸ”­ What am I looking for?

- A fulltime fully-remote role (currently living in Thailand 😎) as an AI fullstack SW developer, integrating AI into the product and SDLC

- Joining a strong team / organization that tackles real-world customer problems at scale with AI

- Working with the best teammates who love learning, adapting, sharing, improving and winning together

πŸ• What's my career timeline?

- 2009 - 2014: Technion - studying for B.Sc in Computer Science

- 2010 - 2019: Rafael - building realtime SW for advanced defence systems (mission-critical, disposable HW, no 2nd chance...)

- 2019 - 2023: AWS - building realtime SW for EC2 Networking devices (world-scale traffic, hard constraints, complex SLDC)

- 2023 - 2024: 1-year break - relocating to Thailand with my wife & 3 daughters

- 2024 - 2026: SaaS - building 2 indie ventures (private repos which I already left), dozens of satisfied paying customers

- 2024 - 2025: Education business - teaching & mentoring non-tech indies about the world of SaaS, > 600 students

πŸ† Achievements

- I'm an ex. AWS SDE, built and deployed prod. SW at the heart of EC2 Networking, world-scale traffic

- My 2nd SaaS involved OCR, which I had no knowledge with, yet we significantly outperformed competitors

- Some students of my education business made their own SaaS ventures and reached paying customers

πŸ’‘ What have I learned?

- Agile & growth mindset are a must and always will be, and so are continuous learning and enthusiasm, AI won't have these soon

- Usually, the 'how' can be solved, the 'what' is just a variation of something known, but the 'why' is what's important, start with 'why'

- SW, like many other things, is a means to a business or personal goal, the goal matters more than whatever SW you build for it

- It's important? Prioritize. It's complex? Document. It's repeating? Automate. It's relevant for others? share. Can AI do it? Let it

- AI can do your work if you teach it, but it's your work and so is the responsibility to lead, direct and validate the AI doing it


Featured Projects

Agent Approval Gate - an LLM email agent that cannot be prompt-injected into sending unapproved email (0/6 breaches vs 3/6 unprotected)
A working proof of concept that the model is not a security boundary. An email-sending agent (TypeScript, Node, Zod, node:crypto) is placed behind a deterministic proxy that validates, canonicalizes and hashes every tool call, classifies it into allow / hold / deny, and holds risky sends for out-of-band human approval bound to the exact arguments by HMAC (single-use, time-bounded). The model has no approval vocabulary at all: the field simply does not exist in its schema. Includes session taint tracking, fan-out rate limits, a hash-chained audit log with a verifier, and an eval harness over a prompt-injection corpus that reports attempt rate and breach rate separately, side by side with an unprotected baseline (3/6 breaches without the gate, 0/6 with it). Seven gate-integrity tests run in CI with no model and no API key; the demo replays recorded model output through the real enforcement path with zero configuration.
Agentic RAG app for SW Incidents Responses - RAGAS Faithfullness 0.950
Built and deployed an agentic RAG system using TypeScript and FastAPI for retrieving info from SW incidents reports and recent deployments with streamed responses. It utilizes both local models (LMStudio) and cloud-based (OpenAI, OpenRouter) that use multiple tools - including web-search (tavily), text-to-sql, subagent delegation and of-course chunks retrieval with hybrid search (pgvector + keyword), metadata filtering and reranking (local / cohere). The system includes baked-in CI (GitHub Actions) and out-of-the-box observability (LangSmith) and evaluations (classic RAGAS, tool selection, chat quality) and reached 0.950 faithfullness in RAGAS score
Claude Code plugin for AI-assisted development, battle-tested and refined across multiple projects.
A collection of Claude Code skills for software development workflows β€” planning, acceptance criteria, execution, code review, research & priming, GitHub bug fixes, git, and validation & reporting
Organizational Agentic Knowledgebase - agents learning from each other, no human involved
A working POC of a shared organizational knowledge base that agents from different users can query from and contribute to in real time, without the human users being aware of this - making institutional knowledge ambient rather than requiring deliberate human curation.

Pinned Loading

  1. ir-copilot ir-copilot Public

    Python 1

  2. ai-dev-env ai-dev-env Public

    Claude Code dev skills collection

    2

  3. ai-dev-env-marketplace ai-dev-env-marketplace Public

  4. autoresearch autoresearch Public

    Forked from karpathy/autoresearch

    AI agents running research on single-GPU nanochat training automatically

    Python 1