Halmar Polanunu
hp·Halmar
PolanunuView CV (English)
Head of Research & DevelopmentEDWi

Strategy.
Products.
People.

I connect people, technology, and business objectives. My experience spans seven years in media intelligence and client solutions at Indonesia Indicator, alongside AI product development, team leadership, and FMCG sales.

Explore my workThe person behind the work
Connecting the dots.A conceptual exploration
ResearchIntelligenceProducts
Business strategyDigital innovationTeam leadership
Scroll to explore

Selected work

Problems.
Decisions. Work.

From understanding what matters to shaping what gets built. A closer look at the work and my part in it.

Product developmentLaunched

Media intelligence application

EDWi

An application that turns data into connected knowledge, retrieves relevant context for questions and tasks, and automatically generates text insights, charts, or complete dashboards. Operators and analysts use its insights to brief executives.

My contribution

I led the team that produced the application, bringing data conversion, a knowledge graph database, contextual retrieval, and automated insight generation into one complete pipeline. As Head of R&D, I set research priorities and technology roadmaps, assess business needs and technical feasibility, and contribute directly to research, prototyping, and testing. I coordinate with technology and business leadership on market priorities and operational handover.

Read the full story

The focus

Help operators and analysts answer questions and perform tasks using relevant context, then deliver insights to executives in a suitable format.

Status & context

The application has launched. It primarily serves operators and analysts, who use the contextual insights to support executive briefings. My wider responsibilities also include developing custom data processing and dashboards for operators, analysts, and executives.

The application pipeline

Connected data. Contextual answers.

  1. Knowledge graph database

    Connect the information

    Convert data into a database that connects entities and the relationships between them.

  2. Contextual retrieval

    Find the relevant context

    Retrieve the information relevant to a question or task, using the connected knowledge.

  3. Automated insight generation

    Produce useful outputs

    Generate contextual insights as text, charts, or complete dashboards.

  • Text insights
  • Charts
  • Complete dashboards

For operators and analysts, with insights escalated to executives.

Simplified workflow of the application; no client data shown.

Indonesia Indicator2019–2025

The decisions
behind the work.

Over seven years at Indonesia Indicator, I supported 20+ client and project portfolios across government, international programs, and private-sector organizations. My roles spanned hands-on analysis, department co-leadership, and strategic analysis. These four studies show selected work from that experience.

Created and ledResearch & testing · 2023–2024

A fluent AI answer still needs an analyst’s judgment.

I created and led work that translated analytical questions into structured AI tasks, then compared model responses with the answers an analyst would expect.

The problem

A model could identify an event and produce a plausible explanation while blurring the difference between an event’s local importance and its relevance to the audience. Correct-looking fields could hide a weak analytical conclusion.

Why it mattered

If the connection to the audience is assumed rather than explained, a well-written output can point attention in the wrong direction.

The review question: what connects this event to the audience we are assessing?

Reconstructed example

Where the analyst challenges the draft

Fictional input: an overseas port schedules maintenance at one terminal. No effect on Indonesian shipping routes has been established.

Model-style draft

The maintenance creates a major risk for Indonesian trade.

Analyst review

The terminal maintenance is the reported event. Relevance to Indonesia remains unestablished: check affected routes, cargo, and alternatives before assigning impact.

Reported factUnproven connectionEvidence to check
The change is in the reasoning: separate the reported fact from the inferred impact, then name the evidence still needed.

Fictional content, created for this portfolio. It demonstrates the method, not an actual product result.

How I approached it

  1. Define the analytical task before writing the prompt.

    I specified the event, category, affected area, and explanation the analysis needed to contain. That gave the test a concrete target beyond a fluent paragraph.

  2. Compare the response with an analyst-written expectation.

    My testing materials placed model responses alongside suggested outputs. This made differences in classification, relevance, and reasoning visible for discussion.

  3. Treat category consistency as part of the method.

    I mapped extraction and categorization steps and highlighted the need for stable categories. Comparable labels matter when individual analyses are brought together.

What the work produced

  • Structured prompt tests and side-by-side output comparisons.
  • Suggested analytical outputs and workflow diagrams connecting extraction, categorization, and review.
Evidence and scope

Archive basis

Prompt-testing materials from 2023, an analytical workflow deck, and risk-analysis framework materials from 2024.

Scope of this case

This case shows framework design, testing, and analytical review. The archive does not establish a measured accuracy improvement or production adoption.

CreatedUser feedback & requirements · 2025

Turn “this is wrong” into a question a team can investigate.

I created feedback and requirements documentation that separated concerns about information selection, analysis, consistency between views, and update timing.

The problem

Users questioned why information outside their criteria appeared, why analytical details did not reflect their reference material, and why different views showed inconsistent information.

Why it mattered

A broad request to improve accuracy leaves several different problems mixed together. Each concern needs its own question and evidence.

Start by identifying which part of the experience the user is questioning.

Reconstructed example

From a user concern to a reviewable requirement

User concernRequirement to discussProposed check
“This item is irrelevant.”Explain how an included item matches the agreed criteria.Inspect an example against the criteria.
“The views disagree.”Keep the same event’s core details consistent across views.Compare the same event in each view.
“How current is this?”Distinguish source publication time from availability in the product.Compare the relevant timestamps.
This sheet reconstructs how I would express and test the concerns. The checks shown here are illustrative proposals, not a record of implemented fixes.

Fictional content, created for this portfolio. It demonstrates the method, not an actual product result.

How I approached it

  1. Separate relevance from the analysis that follows.

    I documented concerns about which items were selected separately from concerns about how selected items were assessed. That distinction gives each issue a clearer place in the discussion.

  2. Describe the inconsistency at the point users saw it.

    I documented where the same information differed across views. Identifying the location of the discrepancy made the reported behavior more specific for investigation.

  3. Make timing and knowledge needs explicit.

    I recorded questions about publication-to-display delay and how uploaded reference documents should inform the assistant. These remained requirements and questions for further work.

What the work produced

  • A structured user-feedback record covering selection, assessment, view consistency, and timing.
  • Document-knowledge requirements and questions for product and technical discussions.
Evidence and scope

Archive basis

User-feedback and document-knowledge notes from 2025.

Scope of this case

My confirmed role is creating the documentation. This case does not claim that I led engineering or that the requested fixes were shipped.

Analysis, writing & team reviewMedia & social intelligence · Selected work, 2019–2023

Client intelligence shaped around the questions that matter.

At Indonesia Indicator, I worked with large volumes of news and social media data to produce insights tailored to each client’s requirements. My contribution combined direct analysis and writing with team leadership and review, depending on the assignment.

The problem

Clients needed to make sense of extensive coverage across platforms and reporting periods. They needed clear explanations of emerging issues, recurring concerns, and the perspectives shaping a story.

Why it mattered

A coverage total alone cannot explain what deserves attention. Reports need to distinguish overall activity from specific concerns and make the reasoning clear to the reader.

Start with the client’s question, then choose the analysis and format that help answer it.

Inside the reporting process

Three ways to make the information useful

Explore original demonstrations of the analysis, structure, and delivery behind the work.

Illustrative examples. All figures, labels, and scenarios are invented. Original client data and report pages are not reproduced.

Analytical report studies02 / 03

Fictional example

Different outlets tell different parts of the story

A subject-by-outlet view helps compare coverage priorities without treating every mention as the same signal.

Invented coverage levels by theme. This describes attention, not sentiment or factual accuracy.
OutletOperationsCommunityEnvironment
Outlet AFrequentLimitedNone
Outlet BLimitedFrequentRegular
Outlet CRegularLimitedFrequent
Outlet DLimitedRegularFrequent
Outlet EFrequentRegularLimited

Review representative articles alongside the matrix to explain framing and whose perspectives appear.

Analytical report studies03 / 03

Fictional example

A short brief makes the open questions visible

Fictional scenario: two neighboring economies announce a pilot process for cross-border trade.

Reported development

A joint statement announces a pilot. Launch timing and eligibility details remain unspecified.

Whose position is being reported?
  1. Public authority

    Emphasizes coordination and implementation.

  2. Industry association

    Requests clearer eligibility and operational guidance.

  3. Community representative

    Raises questions about local access to the pilot.

Next briefing question

What guidance has been published, and which questions remain unanswered?

Condense developments, attribute positions, and distinguish the reported facts from the analyst’s interpretation.

Delivery follows the client’s needs

Daily · Weekly · Monthly · Quarterly · Half-yearly · Annual

Custom application dashboards and reports, presentation decks, or Excel reporting.

My contribution

A combination of hands-on analysis, writing, team leadership, and review, depending on the assignment.

How I approached it

  1. Organize large volumes around meaningful questions.

    Group news and social media coverage by issue, source, and relevant actor, then examine the stories behind the totals.

  2. Match the time frame to the reporting need.

    Produce daily, weekly, monthly, quarterly, half-yearly, and annual insights based on the client’s requirements, with focused reports for specific issues.

  3. Adapt the delivery to the client’s workflow.

    Present findings through dashboards and written reports within custom applications, presentation decks, or Excel. My involvement in analysis, writing, leadership, and review varied by assignment.

What the work produced

  • Recurring intelligence reports and focused briefs explaining issues, coverage trends, and relevant perspectives.
  • Dashboards, application-based written reports, presentations, and Excel reports tailored to client requirements.
Evidence and scope

Archive basis

Selected annual social listening, monthly media mapping, foreign-affairs briefings, and special-report examples from the analyst period.

Scope of this case

My contribution combined hands-on analysis, writing, team leadership, and review, depending on the assignment. Reports could involve colleagues. The examples above are newly created fictional demonstrations and do not imply sole authorship of the source reports or a confirmed client decision.

LedDepartment co-leadership · 2021–2023

Plan around the effort and timing behind each report.

As International Department Co-Leader, I co-led a team of 10 media and data intelligence analysts and led work on workload planning and reporting operations, connecting delivery requirements with available capacity.

The problem

Reports differ in frequency, timing, complexity, language needs, and preparation effort. Counting assignments does not describe when the work must happen or how much attention it requires.

Why it mattered

A plan can look balanced by report count while placing several demanding deadlines in the same part of the day.

Capacity is a question of effort and timing together.

Reconstructed example

Why equal report counts can hide unequal work

AssignmentTiming constraintPlanning question
Recurring briefFixed delivery windowWho is available when it is due?
Longer reviewMore preparation and synthesisWhere is the preparation time reserved?
Additional requestArrives outside the recurring planWhat existing work must be reassessed?
A fictional planning example. It uses qualitative workload differences rather than internal staffing figures.

Fictional content, created for this portfolio. It demonstrates the method, not an actual product result.

How I approached it

  1. Bring the work into a common planning view.

    I guided work that considered report types, schedules, effort, and complexity. The planning discussion needed to account for differences between recurring and additional requests.

  2. Use the workload view to discuss capacity needs.

    My leadership role connected reporting demand with the question of how much analyst capacity was required. The archive includes planning material for that discussion.

  3. Keep the limits of the plan visible.

    The planning materials noted coverage gaps, including some special and longer-cycle reports. Those limits matter when interpreting the capacity picture.

What the work produced

  • Workload-planning discussions supported by report schedules, effort categories, and capacity-planning materials.
  • Operational direction and quality control across daily through annual reporting cycles, alongside mentoring analysts and supporting new client pitches.
Evidence and scope

Archive basis

Workload-planning materials from 2022, department-service planning from 2023, and the updated CV confirming a team of 10 analysts and the co-leadership responsibilities.

Scope of this case

My contribution here is leadership. The CV records streamlined monitoring and reporting workflows to reduce delivery-task time; this case does not assign a numerical improvement or claim personal authorship of every workbook.

The examples are newly written reconstructions. They illustrate my approach without reproducing client data, internal prompts, or product interfaces.

Still asking.
Still building.

Current development directions at EDWi.

01

Connected knowledge

Developing a knowledge graph to connect information and support organizational knowledge management.

In development
02

Issue detection

Developing capabilities to identify issues across news and social media.

In development
03

AI-assisted content

Developing a system in which multiple AI agents collaborate on news and social media content creation.

In development

Different worlds. One connecting thread.

A wider
perspective.

Business and product strategy, cross-functional delivery, and commercial relationships. Seven years at Indonesia Indicator included support for 20+ client and project portfolios across government, international programs, and private-sector organizations.

View CV (English)

Jan 2026 – Present

Head of Research & Development

PT Eka Data Wijaya

Led the team that produced a media intelligence application with a complete pipeline from data conversion into a knowledge graph database to contextual retrieval and automated insight generation as text, charts, or complete dashboards. Leading AI researchers and contributing directly to research, prototyping, and testing. Developing custom data processing and dashboards for operators, analysts, and executives. Setting research priorities and technology roadmaps, assessing business needs and technical feasibility, and coordinating development and operational handover.

Oct 2023 – Dec 2025

Strategic Analyst

PT Indonesia Indicator

Translated client requirements into custom data processing and dashboards for operators, analysts, and executives, including tailored AI-enabled data solutions. Worked with clients, analysts, and technical teams from requirements and design through implementation and refinement. Used news and social media analysis to inform advice on risks, crisis management, and strategic planning.

Jan 2021 – Sep 2023

International Department Co-Leader

PT Indonesia Indicator

Co-led a team of 10 media and data intelligence analysts, coordinating client projects and quality control across daily through annual reporting cycles. Streamlined monitoring and reporting workflows to reduce time spent on delivery tasks, mentored analysts, and supported new client pitches. Concurrent with my analyst role.

Jan 2019 – Sep 2023

International Media Analyst

PT Indonesia Indicator

Used news and social media monitoring tools extensively to track coverage, sentiment, and emerging issues. Produced daily, weekly, monthly, quarterly, and annual intelligence reports on reputation, risks, and opportunities for government and private-sector clients.

Sep 2017 – Mar 2018

District Sales Manager

PT Coca-Cola Amatil Indonesia

Managed relationships with more than 200 modern trade partners across five East Java territories and led merchandisers. Earlier role: Graduate Trainee, Mar–Aug 2017.

Presenting to a seated audience.

A little more human.

Curiosity is
the through line.

My experience brings together media intelligence, client advisory, digital product development, and commercial leadership. At Indonesia Indicator, I supported 20+ client and project portfolios over seven years and co-led a team of 10 media and data intelligence analysts. Earlier, I managed B2B relationships with more than 200 modern trade partners across five East Java territories.

Today, as Head of R&D, I lead AI researchers while staying directly involved in the work. My focus connects business and product strategy, stakeholder management, process improvement, and practical AI-enabled solutions. My academic background is in international relations and political science.

Education

Master’s in International Relations

Universitas Indonesia · Feb 2025 · GPA 3.51 / 4.00

Bachelor’s in Political Science

Universitas Airlangga · Oct 2016 · GPA 3.46 / 4.00

Tools & practice

News & social media monitoring · Power BI · Excel · PowerPoint · Word · n8n · AI coding agents · Prompt engineering

Indonesian: native · English: professional working proficiency

South Tangerang, Indonesia

The next conversation.

Back to top

Let’s connect
the next dots.

For conversations about leadership roles in business strategy, digital innovation, products, research, or intelligence.

LinkedIn profile