Skip to content
View teckedd-code2save's full-sized avatar
💭
Learning, building, evolving
💭
Learning, building, evolving

Block or report teckedd-code2save

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
teckedd-code2save/README.md

Edward Twumasi

Distributed systems · Agent infrastructure · Applied AI

I’m a Lead Backend Engineer at Hubtel and founder-engineer at Serendepify. I build stateful backend systems, agent infrastructure, and low-resource-language AI, with a bias toward the parts that fail in production: orchestration, concurrency, identity, deployment, evaluation, observability, and recovery.

Portfolio · Research · Writing · LinkedIn

Current systems work

Project The engineering problem Explore
GroundControl An agent-operable infrastructure control plane for applications on operator-owned servers. Remote agents use scoped MCP/OAuth actions instead of generic shell access; mutations become durable idempotent operations; connector capabilities are verified independently; exact deployed revisions can be reproduced in ephemeral sandboxes before repair. Code + architecture · Host terminal design
RentAWeekend A real-world agent execution system that turns uncertain conversational intent into research, saved choices, reviewed arrangements, payments, and human-executed work without silently crossing effect boundaries. Recent work includes row-lock/uniqueness protection for retries, transactional provenance, provider shadowing, and checkpointed recovery from model/provider failures. Live product · Portfolio
Ghana Health AI A Twi-first speech-and-language research stack connecting corpus provenance, ASR, human review, model adaptation, semantic evaluation, GPU training, and a deployed voice product. Model promotion is gated by evidence rather than demo quality. Code · Research · Models
Backend as Natural Language A Rust research runtime for lowering controlled natural-language backend declarations into typed intermediate representations, deterministic plans, validation traces, and executable capabilities. Portfolio overview

What I tend to work on

  • Distributed and durable backends: C#/.NET, Temporal workflows, Kafka, Akka.NET, PostgreSQL, Redis, Elasticsearch/OpenSearch, failure handling, concurrency-sensitive financial and public-sector workflows.
  • Agent infrastructure: scoped identity and authorization, long-running operations, idempotency, sandboxed reproduction, human approval boundaries, deployment/recovery tooling, MCP.
  • ML systems and evaluation: Twi/Akan speech and meaning, corpus lineage, GPU training on Modal, model promotion gates, negative-result retention, and production inference.
  • Developer infrastructure: Docker/Compose, Linux, Caddy/Nginx, GitHub Actions, Azure DevOps, cloud deployment, observability, and self-hosted control planes.

Engineering standard

I’m interested in systems where the architecture has to survive retries, partial failure, unreliable model output, human approval, and production operations. I try to make the evidence inspectable: PRs, tests, evaluation records, deployment identities, benchmarks, failure traces, and explicit limitations.

Research

My current research centers on low-resource Ghanaian language systems, especially Twi speech and semantic recovery. I keep research artifacts, reviewed data, model candidates, and production promotion as distinct evidence classes.

I also co-authored a peer-reviewed paper in African Geographical Review on geocoding polling-station data and accessibility in Ghana.

Work with me

I’m especially interested in distributed infrastructure, agent/research platforms, ML systems, and teams building AI that has to operate reliably outside a notebook.

Explore the work · Read the engineering notes · Get in touch

Pinned Loading

  1. ghana-health-ai ghana-health-ai Public

    Voice-first Ghana Health AI — Twi maternal health, ASR, market ecommerce

    TypeScript

  2. groundcontrol groundcontrol Public

    GroundControl is a self-hosted infrastructure dashboard for managing Docker containers, Caddy proxies, deployments, and system health on a Hetzner-class VPS

    TypeScript 1

  3. convoy convoy Public

    The deployment agent that ships your code — not a rewrite of it.

    TypeScript 1

  4. akan-speech-lab akan-speech-lab Public

    Speech AI research lab for the Akan language — building ASR, TTS, and voice datasets for one of West Africa's most spoken languages.

    Python

  5. integration-skills integration-skills Public

    Agent-first TypeScript integration skills

    JavaScript

  6. ai-build-tools ai-build-tools Public

    From product spec to live URL — scaffold a backend with Forge, ship to your VPS with ship-to-vps

    TypeScript 2