Skip to content
View adityonugrohoid's full-sized avatar

Block or report adityonugrohoid

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 supported. This note will be visible to only you.
Report abuse

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

Report abuse
adityonugrohoid/README.md

Hi, I'm Adityo Nugroho

18+ years in Network Performance & Operations, transitioning into AI/ML Engineering.

After managing critical network operations serving millions of users for 18 years, I learned what breaks at scale. I pivoted into AI/ML engineering to address these bottlenecks, building and operating 12 systems over 14 months, spanning data engineering, model deployment, and agentic automation.

 

Featured Projects

Autonomous Network Operations: Observe -> Decide -> Act

A cohesive production-grade suite solving the "Mean Time to Recovery" bottleneck.

  • Incident Commander: Async log analyzer using Gemini 2.0 Flash Lite to process 500 logs/sec. Reduces 3,000 raw logs to 1 incident report (63x noise reduction).
  • NOC-Oracle: RAG-powered troubleshooting assistant with Gemini 2.0 Flash. Uses hybrid search (Vector + Keyword) to achieve 100% retrieval accuracy and 0% hallucinations.
  • Net-Ops Agent: Agentic AI using Gemini 2.0 Flash Function Calling. Enforces 100% human-in-the-loop approval for all operational actions.

Production Algorithmic Trading Bot

  • Status: Live 24/7 on Cloud VPS.
  • Tech: Python Async, WebSockets, Systemd, Ed25519 Auth.
  • Strategy: Proprietary dynamic trailing take-profit with exponential decay and regime-aware auto-compounding.
  • Architecture: Handles real-time market data with sub-100ms latency. Orchestrated via systemd for automatic recovery. Proves ability to ship and maintain always-on infrastructure.

End-to-End Pipeline: Synthetic Generation -> ML Analytics

A complete data science ecosystem combining physics-based simulation with rigorous ML.

  • Digital Twin: Physics-based generator producing 5.6M sessions with SINR/latency modeling. 100% reproducibility.
  • QoE Analytics: ML pipeline (XGBoost/LightGBM) achieving R-squared 0.9997 (driven by highly correlated physics settings) and identifying congestion bottlenecks (Cohen's d = -2.75).

Other Notable Work

  • RATU Trading Suite: 4-component crypto infrastructure (Multi-chain Scanner, On-Chain Holder Analytics, Market Data REST API, FIX Bot).
  • AI Studio: Generative AI interior design app (Next.js 14 + Gemini Multi-Model) delivering photorealistic renders from architectural sketches.
  • Telecom ML Framework: Open-source v1.0.0 framework for telecom data science covering 6 use cases (Churn, RCA, Anomaly, QoE, Capacity, Optimization).

 

Technical Stack

Domain Technologies
Architecture Python Async, Systemd, Linux VPS, Event-Driven, Microservices
AI Models Google Gemini 2.0 Flash/Lite, Gemini 2.5 Flash Image, Gemini 3.0 Pro Image Preview
AI Engineering RAG (ChromaDB), LangChain, Agentic Patterns, Pydantic, Function Calling
ML & Data XGBoost, LightGBM, SHAP, Pandas, Parquet, Synthetic Data (SDV)
Protocols REST, WebSockets, FIX 4.4, JSON-RPC, GraphQL
DevOps Docker, CI/CD, Pytest, Ruff, Mypy, UV Manager

 

What I Bring

  • 18+ Years of Domain Authority: Operating large-scale networks: I've lived the pain points, not guessed them. I bring this experience into the systems I now automate.
  • Real-World Design: My systems handle actual constraints: partial outages, race conditions, and rate limits. "Happy path" code does not survive production.
  • Measurable Business Impact: I translate technical complexity into outcomes that matter: 12 production-grade systems architected and deployed in 14 months.

 

Available For

  • Observability & APM: Automated remediation and log analysis.
  • Fintech & Trading: Low-latency decision systems.
  • Industrial AI: Digital twins and predictive maintenance.
  • Location: Remote Preferred | Based in Indonesia (UTC+7)

 

Connect

Pinned Loading

  1. incident-commander incident-commander Public

    Asynchronous log analyzer using Gemini 2.0 Flash Lite, reducing 3,000 raw logs to a single incident report (63x noise reduction).

    Python

  2. noc-oracle noc-oracle Public

    RAG engine using Gemini 2.0 Flash with hybrid search, achieving 100% retrieval accuracy.

    Python

  3. net-ops-agent net-ops-agent Public

    Agentic AI using Gemini 2.0 Flash Function Calling, enforcing 100% human-in-the-loop approval.

    Python

  4. trailing-edge trailing-edge Public

    Async Python trading bot for Binance with trailing take-profit, Donchian channels, regime detection & Ed25519 auth

    Python

  5. telecom-digital-twin telecom-digital-twin Public

    Deterministic synthetic telecom data generator with physics-based network KPIs. Produces multi-table LTE datasets (users, cells, sessions, events) for ML/analytics practice.

    Jupyter Notebook

  6. telecom-qoe-analytics telecom-qoe-analytics Public

    End-to-end Data Science portfolio: EDA, statistical testing, ML modeling (XGBoost, LightGBM), and anomaly detection on telecom QoE data. Six-phase analytics pipeline.

    Jupyter Notebook