Skip to content
View adhikasp's full-sized avatar

Organizations

@technocorner @coala @pipe-dream

Block or report adhikasp

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. 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
adhikasp/README.md

Adhika Setya Pramudita

I'm a software engineer working on applied AI and large-scale backend systems. At Grab, I lead AI Enablement—building company-wide agent infrastructure, MCP tooling, and internal AI workflows across engineering and business teams. Previously, I worked on the GrabMerchant Assistant, MerchantOps, and real-time merchant data pipelines.

Outside of work, I build tools that make coding agents and personal software easier to steer, run small public data experiments, and self-host what I use daily.

Current focus

  • Coding agent harnesses & supervision: making agent execution observable and steerable instead of trusting blind shell exit codes or long system prompts (agshell, steer).
  • Model Context Protocol (MCP) tooling: lightweight CLI clients and servers that wire LLMs directly into terminal workflows and external data sources.
  • Public data & notebooks: maintaining a daily-updated Parquet mirror of every Hacker News item since 2007 and querying it with DuckDB to see how tech conversations shift over time.

Selected work

Agent & MCP tooling

  • mcp-client-cli — Terminal client for running LLM prompts against any Model Context Protocol server across OpenAI, Anthropic, Groq, or local models.
  • agshell — Supervised bash-compatible shell in Rust (built on brush) that reports per-command and per-pipeline-stage status as typed JSON instead of a single exit code.
  • steer — Per-turn rule injection for Claude Code and OpenCode using a decision model to pick the single harness rule relevant to what the agent is about to do.
  • MCP servers — Data connectors for GitHub repositories, Reddit, Twitter/X, YouTube transcripts, and AccuWeather.
  • herdr-command-plugin & herdr-app-mods — Command palette (Cmd+K), fuzzy switcher, and auto-naming tabs for managing multi-agent terminal sessions in Herdr.

Apps, data & simulations

  • hn-data-fetcher — Async fetcher that streams Hacker News items straight into Hive-partitioned Parquet files, powering the public HN dataset.
  • brainmax — Vertical-swipe feed for arXiv, bioRxiv, medRxiv, PubMed, and HN with LLM-simplified abstracts and daily animated SVG explainers (Rust + SolidJS; live).
  • kodepos — Interactive map visualizing how Indonesia's five-digit postal code hierarchy maps onto actual geography (live).
  • mechanical-clock — Wall clock in Rust and Rune whose hands are driven by a simulated gear train and a 5-beat/s lever escapement rather than animation timers.

Recent writing on adhikasp.id

Stack

  • Languages: Go, Python, Rust, TypeScript, SQL
  • AI & data: LLM evaluation & agent harnesses, MCP, DuckDB, Parquet, Apache Druid, Kafka
  • Systems & infra: Linux, Docker, Kubernetes, Terraform, AWS, GCP, self-hosted services

Elsewhere

Pinned Loading

  1. mcp-client-cli mcp-client-cli Public

    A simple CLI to run LLM prompt and implement MCP client.

    Python 676 82

  2. mcp-youtube mcp-youtube Public

    Model Context Protocol to fetch youtube transcript

    Python 50 20