Graduate Teaching Assistant
Carnegie Mellon University, Pittsburgh
- TA for 90-710 Applied Economic Analysis and 94-844 Generative AI Lab
Rizaldy Al Kautsar Utomo
AI strategist and data scientist. I turn machine learning and maps into products and policy that hold up in the real world.
- Now
- MS Public Policy & Management, Data Analytics at Carnegie Mellon Heinz College (GPA 3.94), and TA for Generative AI Lab
- Before
- 7 years at Telkomsel, Southeast Asia's largest telco with 170M users. Built a data team that earns USD 11M a year.
- This summer
- Built a geospatial monitoring platform for Pittsburgh's Urban Redevelopment Authority that replaced a $15,000-a-year subscription
- Where
- Pittsburgh, PA, by way of Jakarta and Bandung
Open to AI safety, geospatial intelligence and solution architecture roles. Remote projects now, full-time from 2027.
AI strategist and data scientist with 7+ years of industry experience translating machine learning, geospatial analytics and large-scale data products into measurable business and policy outcomes.
Experience
Geospatial Intelligence Intern
Urban Redevelopment Authority, Pittsburgh
- Engineered an in-house geospatial monitoring platform to replace a paid survey subscription, cutting $15,000 in annual licensing and bringing contractor field reporting under URA control
- Shipped an executive KPI dashboard, map and contractor portal on ArcGIS
- Pioneered agentic development end to end and handed over an agent-ready repository so the GIS team can keep shipping without a dedicated engineer
Data Scientist II
Telkomsel, Jakarta
- Deployed econometric models for the Telkomsel Emergency Package loan product, achieving a 6% revenue increase and contributing to an all-time-high revenue milestone
- Built an e-commerce install predictor from telco features with CatBoost, yielding 2.5x lift over a random baseline
- Led data engineers to move the Mobility Data Pipeline from yearly to monthly, cutting delivery from 4 days to 16 hours
- Led a multinational Singtel Group team to win Action for Change 2024 with EcoFlow, a geospatial-data venture
Associate Data Strategist
Telkomsel management trainee, Jakarta
- Built the data team for Consultative Insight, a new line selling aggregated data to ride-hailing clients, generating USD 11M in annual revenue
- Rebuilt programmatic advertising segmentation from browsing-interest data, across 51 interest segments aligned with international standards
Market Intelligence Analyst
Lazada Indonesia, Jakarta
- Ran a data-driven promo campaign for younger shoppers, lifting CTR from 0.5% to 4%
- Replaced formula lookups with SQL inside Google Sheets for faster SKU search and inventory management
Education
Carnegie Mellon University, Heinz College
MS Public Policy & Management, Data Analytics. Concentration in AI Management. GPA 3.94.
- Generative AI Lab, Machine Learning for Public Policy, Responsible AI
Institut Teknologi Bandung
BS Information Systems & Technology. GPA 3.55.
- Probability and Statistics, Enterprise Architecture, Database Modelling
Projects
BukuGambar.AI
- A web app that turns text or photos into coloring-page outlines with the OpenAI API
EcoFlow, founder and team lead
- AI-powered sustainable tourism venture that uses telco data to plan crowd-aware itineraries. Incubated by Telkomsel's innovation unit and now live as Telkomsel Travel Assistant.
tSurvey, research and insight
- Built an audience-sizing engine and stratified sampling framework over 158M respondents
- Engineered an AWS OCR pipeline that turns paper surveys into structured data
Awards and certifications
- Global participant, Build with Opus 4.7 Hackathon, invite-only (Anthropic, 2026)
- 2nd place, Pittsburgh AI Safety Hackathon, policy analysis of AI-enabled mass surveillance (Carnegie Mellon, 2026)
- 1st place, Prompt-a-Thon, agentic AI ideation (Microsoft Indonesia, 2025)
- LPDP Scholarship, full ride for a master's degree (Indonesia Endowment Fund, 2024)
- 1st place, AWS PRFAQ Challenge (Amazon and Telkomsel, 2023)
- Ganesha Karsa, ITB's highest distinction for multiple competition wins (2018)
- Generative AI, Google Cloud (2024); AWS Technical Essentials (2018)
Skills
- Programming
- Python, SQL, PySpark, HTML, JavaScript
- Agentic development
- Multi-agent orchestration, harness design, MCP, Claude Code, Codex, CI/CD, RAG, model fine-tuning
- GIS
- ArcGIS Pro, QGIS
- Data visualization
- Tableau, Power BI, Looker, Streamlit, Plotly, Seaborn, Matplotlib
- Databases
- PostgreSQL, MySQL, Hadoop (Cloudera), MongoDB
Things I built for people I care about
Ventures
Carnegie Mellon coursework
Competitions
BukuGambar.AI
OpenTurns text or a photo into a coloring page. Built for my sons, used by other parents too.
Distance to CMUShare your location to measure
StatusShowing Pittsburgh
Your location stays in your browser. Nothing is stored or sent anywhere.
Accessible information is not a feature, it's the whole point. When machines extend what humans can do, something has to steer. Let it be kindness, not engagement metrics, not extraction.
Build for the person with no second option. Question every model, especially your own. Make technology that can be explained to the people it affects. Ship things that work, and let the impact speak.
Rizaldy, 2026
- Read the resume
- Open a project
- Measure your distance to CMU
- Look in the Trash
- Find "About allrize…"
0 of 5 done
Named after the tuyul, a small spirit from Javanese folklore that quietly collects money for its keeper. Here, the spirit is a team of agents.
| Book | Return | Tuyul's lead |
|---|---|---|
| Tuyul agentic desk | +14.51% | — |
| S&P 500 (SPY) | +5.53% | +8.98 pts |
| IHSG (Jakarta Composite) | +4.16% | +10.35 pts |
Real money, small book. Tuyul's return is time-weighted in USD, so deposits don't count as gains; IHSG is in rupiah. Past results say nothing about the future, and this isn't investment advice.
- 1. Scheduler wakes the desk
A few timed sessions each market day, plus a Monday deep scan and a Friday retro.
- 2. Load memory
A context pack pulls from the second brain before anyone acts.
Obsidian vaultnotes and researchStrategy + decisionsthe constitution, every change of mindJournalswhat happened last time - 3. Agents deliberate
Each agent has one job, and they check each other.
Lead agentClaude runs the session as investor and operatorCouncila second model does research and pushes backScorertyped scores for regime, momentum and risk - 4. Guardrails decide what's allowed
Kill switch, per-trade dollar cap, take-profit and stop-loss on every order, exposure caps by market regime. Rebalances and sells wait for a human yes.
- 5. Act through MCP
The lead agent calls broker and market-data tools over MCP: preview first, then place.
- 6. Write it all down
Every decision lands in the repo as a journal entry, and CI audits it. The Friday retro feeds lessons back into memory, then the loop starts again at step 1.
Why it matters: it's the same pattern I bring to work. Agents with narrow roles, shared memory, hard limits and a paper trail a human can audit.
- Yearly mobility pipeline
Rebuilt to run monthly. Delivery went from 4 days to 16 hours. Telkomsel.
- $15,000/yr survey subscription
Replaced by an in-house geospatial monitoring platform. Pittsburgh URA, 2026.
- 0.5% click-through rate
Retargeted the promo to younger shoppers. It hit 4%. Lazada, 2017.
- Stacks of paper surveys
An AWS OCR pipeline turned them into structured data. tSurvey.
- VLOOKUP spaghetti
Swapped for SQL inside Google Sheets, so SKU search got fast. Lazada, 2017.