Skip to content
View Dandiggas's full-sized avatar
20 followers · 13 following

Achievements

Achievement: YOLOAchievement: Pull Sharkx2Achievement: Pair Extraordinaire

Achievements

Achievement: YOLOAchievement: Pull Sharkx2Achievement: Pair Extraordinaire

Block or report Dandiggas

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.

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
Dandiggas/README.md

Hi 👋 I'm Daniel Adekugbe

Software engineer at the intersection of platform infrastructure, security, and AI, making agents safe to run in production.

I build and operate the controls that let companies hand real power to AI agents: per-agent identity and least-privilege tool scoping, gated actions ("propose, never promote"), full trace audit, and eval-backed verification. Day job at Sophos: a production agent workflow platform and MCP servers over a threat-intelligence graph. The same patterns run in the public working code below.

I come at this from infrastructure and security, not ML: Python and Go services on AWS (EKS, ECS, Lambda, SQS, IAM), Terraform and CDK, Kafka, CI/CD, structured observability.

Selected work

  • threat-intel-brief - AI-reliability harness: the LLM is confined to one graded boundary, output is grounded and citation-checked, golden eval set with a regression gate in CI
  • recruiter-risk-triage-v1 - local-first agentic dashboard for safely triaging recruiter outreach: verification evidence, containment of risky assessments
  • shadow-ai-scout - autonomous due-diligence agent for shadow-AI vendor risk: plan/act/observe/verify loop, rejects any claim without a source quote, full audit trail
  • claude-session-memory - published Claude Code skill governing agent memory lifecycle: typed memories, propose-and-persist, correction capture
  • studio-agent-daw - natural-language control of DAWs (Ableton, Reaper, Logic) via MCP servers I build, not just consume
  • The Shed (MusiciansPracticeApp) - shipped browser practice workbench for musicians

Principles I build by

  • Curated, parameterised tools over raw passthrough. An agent is an injection surface.
  • Propose, never promote. Agents draft, humans commit.
  • Every action attributable: per-agent identity, scoped tokens, audit trails.
  • An LLM step earns its place with evals and a regression gate, or it ships as deterministic code instead.

Writing lives at tech.dandiggas.com. I'm also a session musician; the music side lives at dandiggas.com.

🌐 Socials:

LinkedInTikTok

💻 Tech Stack:

Languages: PythonGoRustJavaScript

Cloud Providers: AWS

Infrastructure as code: TerraformCloudFormation

Containerisation Tools: DockerKubernetesHelm

CICD: GitHub Actions

Frameworks: FastAPIDjangoDjangoRESTFlaskReactVue.jsNext JS

Database: AmazonDynamoDBPostgres

Pinned Loading

  1. dandiggas-sitedandiggas-sitePublic

    Personal site + engineering blog — AI workflow tools for music and security

    TypeScript

  2. claude-session-memoryclaude-session-memoryPublic

    Persistent memory and context loading for Claude Code sessions

  3. keyboard-manual-assistantkeyboard-manual-assistantPublic

    RAG system for keyboard synthesizer manuals with Langfuse observability

    Python

  4. MusiciansPracticeAppMusiciansPracticeAppPublic

    The Shed: a browser practice workbench for musicians to save sessions, tracks, licks, takes, charts, metronome and tuner context.

    TypeScript 1

  5. recruiter-risk-triage-v1recruiter-risk-triage-v1Public

    Local-first agentic dashboard for safely triaging recruiter outreach, verification evidence, and risky assessments.

    Python