Feb 2025 to nowAI Engineer, CEF Co., Ltd.
I work across robotics, LLM applications and IoT systems, and I also look after the servers the team runs on.
Humanoid robot manipulation
- What I did
- Built a custom grasping environment in HumanoidBench with the Unitree H1 robot and Shadow Hand.
- Operated and supported humanoid robot systems on the research side.
- How it works
The environment defines the robot, the object and the grasp task in simulation, so grasping methods can be tried and compared before touching real hardware. Reinforcement learning frameworks can then train the policies.
- Tech stack
- MuJoCoIsaac LabStable Baselines3PPO and SACUnitree H1
- Impact
- Gave the team a reusable simulation setup for studying humanoid grasping.
Autonomous agricultural robot
- What I did
- Worked on the navigation and mapping software of a farm robot.
- Worked on computer vision for agricultural use.
- How it works
ROS2 connects sensors, planning and motor control as separate programs that talk to each other. SLAM builds a map of the field while estimating the robot's own position in it, which autonomous driving depends on.
- Tech stack
ROS2SLAMComputer vision- Impact
- Contributed to the robot's ability to navigate on its own.
News generation system
- What I did
- Built a system that drafts news articles with an LLM.
- Fine-tuned the model with LoRA.
- How it works
Qwen is the base language model. LoRA trains small adapter layers instead of all the weights, so fine-tuning fits on limited GPU memory. The scraping pipeline supplies the source text.
- Tech stack
- Qwen LLMLoRA
Python
Docker - Impact
- Gave the company an in-house way to produce draft articles automatically.
Data pipeline and web scraping
- What I did
- Built an end-to-end pipeline that collects data from the web, processes it and passes it on.
- Ran it in GPU-enabled Docker containers on Ubuntu.
- How it works
Selenium opens real browser sessions for pages that need JavaScript, BeautifulSoup extracts the content, and Python cleans and stores it. Docker keeps the environment identical across machines.
- Tech stack
Python
SeleniumBeautifulSoup
Docker
Ubuntu- Impact
- Turned scattered web sources into a steady data supply for the team's AI work.
IoT dashboard
- What I did
- Built a full-stack dashboard for data from ESP32 devices.
- How it works
ESP32 microcontrollers send readings over MQTT, a lightweight messaging protocol made for small devices. A Flask backend serves the data and JavaScript draws the dashboard.
- Tech stack
ESP32
MQTT
Flask
JavaScript- Impact
- Turned raw device output into something the team can monitor visually.
Server administration
- What I did
- Administered the Linux servers and Docker setups the team uses.
- How it works
Installing and maintaining services, managing containers, and troubleshooting when something breaks.
- Tech stack
Linux
Docker
Ubuntu- Impact
- Kept a consistent environment available for the team's AI work.
Sep 2023 to Jan 2024Data Analyst, Indonesia Stock Exchange (BEI)
I worked as a data analyst, automating data work with Microsoft Power Automate and building reports and dashboards in Power BI.
Data input, merge and split automation
- What I did
- Built automated processes for data input, merge and split operations with Power Automate.
- How it works
Flows run the repetitive steps in the background, moving data in, combining files and splitting them, so people do not repeat the same copy and paste work.
- Tech stack
- Power Automate
- Impact
- Reduced repetitive manual data handling.
Email reminders and approval workflow
- What I did
- Set up an email scheduler that works as a reminder system.
- Implemented an approval workflow process for task management.
- How it works
Scheduled flows send reminder emails on a timetable. Approval flows route an item to the right person, wait for the answer, and record it.
- Tech stack
- Power AutomateOffice 365 Outlook
- Impact
- Helped tasks get followed up and decisions get recorded.
Document pipeline: email to OCR to Excel
- What I did
- Built a flow that takes initial share information documents received by email, kept in OneDrive, and reads them with OCR in AI Builder.
- Stored the extracted data in Excel.
- How it works
Documents arrive by email and are kept in Microsoft OneDrive. Power Automate picks up each new file and sends it to a custom AI Builder document model, trained on tagged samples, which reads the PDF with OCR and returns the fields. The result is written into Excel, which serves as the storage.
- Tech stack
- Power AutomateAI BuilderOneDriveExcelOCR
- Impact
- Turned emailed PDFs into structured data in one place instead of manual re-typing.
Daily listed report and financial summaries
- What I did
- Created a Daily Listed Report.
- Generated Financial Summaries.
- How it works
Power BI dashboards read from the underlying data so the report and the summaries give an up-to-date view without being rebuilt by hand.
- Tech stack
Power BI- Impact
- Gave stakeholders a current overview of listed companies.
Risk profile visualizations
- What I did
- Developed Risk Profile visualizations for listed companies.
- How it works
Charts in Power BI summarize each company's risk information so differences between companies are easy to compare.
- Tech stack
Power BI- Impact
- Made risk information easier to read at a glance.
Journalist and helpdesk dashboards
- What I did
- Designed a Journalist Dashboard for media-related analytics.
- Established a Helpdesk Dashboard for support operations.
- How it works
Each dashboard is shaped around its audience: media-related figures for the first, support activity for the second.
- Tech stack
Power BI- Impact
- Gave each team a single place to see its numbers.
ArcGIS spatial analysis
- What I did
- Visualized the coordinate data of listed companies in ArcGIS.
- Integrated ArcGIS with Power BI for spatial analysis.
- How it works
The coordinate data stored from the pipeline is plotted in ArcGIS as a map of where the companies are. Connecting ArcGIS to Power BI put that map inside the interactive dashboards, where it can be filtered like any other chart.
- Tech stack
ArcGIS
Power BI- Impact
- Made the geographic distribution of listed companies visible at a glance.
Jul 2023 to Mar 2024Freelance IT Developer, Bakti Kominfo
I built monitoring and data tools around the Palapa Ring fiber optic network.
Fiber optic route dashboard
- What I did
- Built a monitoring dashboard for the Palapa Ring fiber optic cable routes.
- Integrated it with disaster alarms.
- How it works
The dashboard shows the cable routes on a map, and disaster alarm information is brought in so hazards can be seen against the infrastructure.
- Tech stack
- Web dashboardDisaster alarm integration
- Impact
- Helped the team see which parts of the network are exposed when a hazard is reported.
Hazard data pipelines
- What I did
- Built API-based data pipelines for BMKG and PVMBG data.
- How it works
BMKG is Indonesia's meteorology, climatology and geophysics agency, and PVMBG is its volcanology and geological hazard agency. The pipelines fetch their data through APIs and deliver it in a form the dashboard can use.
- Tech stack
- REST APIsData pipelines
- Impact
- Kept hazard data flowing into the dashboard automatically.
Database design and QA
- What I did
- Designed the database architecture.
- Did QA testing.
- How it works
I structured the data so dashboards and pipelines could read it reliably, and tested features before release.
- Tech stack
- Database designQA testing
- Impact
- Gave the project a stable data foundation.
Jan to May 2023Junior Data Processing (contract), NielsenIQ
I processed large datasets and kept their quality in check.
Data processing and validation
- What I did
- Processed large datasets.
- Built automated validation and cleaning pipelines.
- Ran data quality checks.
- How it works
Automated rules catch missing, inconsistent or out-of-range values before the data is used.
- Tech stack
- ExcelVBA
- Impact
- Reduced manual checking effort.
Dashboards
- What I did
- Built dashboards with Tableau, VBA and Excel.
- How it works
Tableau provides interactive visuals, while Excel and VBA automate the reporting around them.
- Tech stack
TableauExcelVBA- Impact
- Let stakeholders read results visually instead of raw tables.
May to Jun 2022Big Data Analytics Intern, PT Kimia Farma Tbk.
I built reports from company data.
Analytics reporting
- What I did
- Built reports with Tableau, SQL, Python and Excel.
- How it works
SQL pulls the data, Python and Excel prepare it, and Tableau presents it.
- Tech stack
- SQL
Python
TableauExcel - Impact
- Made company data easier to read for decision makers.