Plenary 3 · INNOVATHON 2026 21 July 2026 · MPC @ MATRADE · Kuala Lumpur
PLENARY 3 · INNOVATHON 2026

AI Transformation of Reward Culture
& Quick-Win Projects

Experience in participating & organising international AI competitions — and what Malaysian hospital teams can steal from the playbook next Monday morning.

H
Dr Ahmad Hakiim Jamaluddin (Ph.D.) Senior Lecturer · Statistics & Data Science · Department of Mathematics & Statistics · Universiti Putra Malaysia
Jump to the case study Six lessons for INNOVATHON 2026 →
189Teams
605Hackers
5Countries
62Institutions
5Winners
4.42/5Experience
Speaker

A quick word about me

📷 Speaker portrait Replace with assets/images/hakiim.jpg if you have a close-up photo. Otherwise this grey placeholder keeps the layout stable.

Where I come from. Grown up in Kedah, took pure sciences, and got hooked on what math could do for real decisions. Postgraduate life took me to UNSW Sydney; first job life began at UUM.

Education. Ph.D. Mathematics & Statistics, UNSW Sydney (2021–2025) — nominated for the Outstanding Ph.D. Theses Award. MSc Statistics & BSc (Hons) Industrial Statistics at UUM — Chancellor Gold Medal (best overall undergraduate, CGPA 3.87 / 4.00).

Now. Senior Lecturer in Statistics & Data Science at UPM's Department of Mathematics & Statistics, Statistical Data Science Group. Fellow of the UNSW AI Institute (fundamental data-science team) and core member of uDASH.

Hats I wear right now:

  • Organising Chair — International Data Science Challenge (IDSC) 2026.
  • Education Sector Lead — AI Talent Working Group, National AI Office (NAIO), Ministry of Digital Malaysia.
  • AI Research Fellow — Sengkang General Hospital, Singapore (2025).
  • Visiting Lecturer — Universitas Brawijaya (UB), Indonesia (2026).
  • External Expert Reviewer — BS Health Data Science Programme, Health Services Academy, Pakistan (2026).

Loves. Good math, good coffee (flat white, please), evening runs. And TEDx.

The Big Question

Why do some hospital teams ship extraordinary AI — and others quietly stall?

Most teams I meet are not short of ideas. They are short of oxygen: time, scoring rules, a clear winner's table, and someone senior clapping on stage day. So I built that oxygen — first as a contestant, then as a chair.

What stalls

  • No visible winner's table — "we'll see"
  • Rubric shifts after submissions
  • Judges who don't speak clinician
  • Prizes the finance team can't explain
  • No follow-up governance after Day 1

What flies

  • One sharp clinical pain point
  • Rubric published before the race
  • Mixed jury: clinician + statistician + journalist
  • Cash + certificate + a stage the family watches
  • A path to deploy by next quarter
A light joke. "If you build it, they will come." → No. "If you reward it, they will come back — with friends." → That one is closer to the truth.
Framework

The Reward Culture Framework — four pillars you cannot skip

Skip any one of these and the engine stalls within a month. NHS folks — this is your long-standing quality-improvement logic in software form.

01

Clear goal

Pick one clinical pain point. Make it sharp. Everyone knows what they are running towards, and what they are not doing.

SpecificMeasurableTime-boxed
02

Fair rules

Publish the rubric before the race. No mystery weighting, no late surprises. People respect what they can read.

PublicStableAuditable
03

Real reward

Cash, certificate, trophy — yes. But equally important: a stage, a story their mother can read in the newspaper, and time with the Director.

VisibleTangibleCareer-relevant
04

Public recognition

Put it on the national stage. Senior judges, livestream, hospital press. Total visibility = total motivation.

National pressHospital C-suitePatient press
The hospital-grade framing. Translate the four pillars into Bahasa Melayu / KKM-speak and they become matlamat jelas, aturan adil, ganjaran nyata, pengiktirafan terbuka — the same four levers you already use when you onboard a new clinical service line.
Quick Wins

Quick-Win Projects — small, fast, visible, useful

If your AI roadmap lives entirely in a strategy deck, it dies there too. Here is the recipe that actually finishes.

Small

One team, one shelf, one pain point. No 18-month stealth project.

Fast

Live in 2–4 weeks, not 2 years.

Visible

A nurse can touch it on Monday morning. Not a slide in a steering committee.

Useful

It already saves someone 12 minutes a day.

Documented

A one-pager the Director can carry to the Ministry of Health Malaysia.

Why quick wins win. They build trust; they free up budget for the bigger bets; they give junior staff an early win — and they stay; they print stories the hospital can carry up the chain; and they create a track record for next year's grant application. The big AI strategy starts here, not at the whiteboard.
Case Study

IDSC 2026 — "Mathematics for Hope in Healthcare"

The International Data Science Challenge 2026, themed "Mathematics for Hope in Healthcare," is a fully online global challenge hosted by Universiti Putra Malaysia (UPM) in alliance with three Indonesian co-organisers and PERSAMA. It ran for 52 days, end-to-end (25 Feb – 17 Apr 2026).

📷 IDSC 2026 logo / banner Replace with assets/idsc-logo.png when publishing.

Built by

  • Host: Department of Mathematics & Statistics, Universiti Putra Malaysia (UPM)
  • Co-organisers: UNAIR (Surabaya), UNMUL (Samarinda), UB (Malang)
  • Partner: PERSAMA — Persatuan Sains Matematik Malaysia
  • Format: Fully online, two stages, free to enter
  • Tight loop: Registration → Stage 1 → Top 10 → Grand Final on 14 April 2026

Datasets (one pick per team)

  • Hillel Yaffe Glaucoma Dataset (HYGD) — retinal fundus, ophthalmology
  • 12-Lead ECG for Brugada Syndrome — cardiac arrhythmia
  • P300-based BCI (bigP3BCI) — EEG / neural signals
  • Myocardial Perfusion SPECT — cardiac nuclear imaging
Why this matters as a case study. It is fully online, cross-border, free to enter, entirely academician + student-led, judged on real clinical safety, published in the national press — and reusable. Featured in Dewan Kosmik by Dewan Bahasa dan Pustaka (DBP).
By the numbers

IDSC 2026 — the proof, not the promise

Operational diagnostics

MetricValue
Total registered teams189 teams
Total participants (all team-member affiliations)605 hackers
Teams progressed to Stage 1131 teams
Complete final technical projects submitted129 teams
Top 10 finalist teams shortlisted10 teams (~ top 5% elite tier)
Unique global institutions represented62 entities (61 universities + 1 industry org)
Geographic footprint5 countries — 🇲🇾 🇮🇩 🇦🇺 🇮🇳 🇹🇯
Undergraduate-led teams172 teams · 91% of field

Team configurations

Team rosterTeams%Total headcount
4 members8645.5%344
3 members5529.1%165
2 members4825.4%96
Total189100.0%605

Top 10 campus powerhouses (by teams led)

#InstitutionCountryTeams led
1Universitas Airlangga (UNAIR)🇮🇩 Indonesia18
2Universiti Putra Malaysia (UPM)🇲🇾 Malaysia15
3 (tie)Universiti Malaya (UM)🇲🇾 Malaysia14
3 (tie)Universitas Negeri Surabaya (UNESA)🇮🇩 Indonesia14
5Universitas Singaperbangsa Karawang (UNSIKA)🇮🇩 Indonesia13
6Universitas Brawijaya (UB)🇮🇩 Indonesia11
7Universiti Kebangsaan Malaysia (UKM)🇲🇾 Malaysia9
8Universitas Mulawarman (UNMUL)🇮🇩 Indonesia7
9 (tie)Universiti Sains Malaysia (USM)🇲🇾 Malaysia5
9 (tie)Universitas Gadjah Mada (UGM)🇮🇩 Indonesia5

Dataset battlefield

Dataset & clinical focusTeams%
Hillel Yaffe Glaucoma Dataset (HYGD) — retinal fundus9751.3%
12-Lead ECG for Brugada Syndrome — cardiac arrhythmia9751.3%
P300-based BCI (bigP3BCI) — EEG / neural signals4423.3%
Myocardial Perfusion SPECT — cardiac nuclear imaging3920.6%

Participant satisfaction (out of 5)

4.42

Overall experience

4.46

Workshop helpfulness

4.25

Format organisation

3.99

Confidence in noisy data

96% of all respondents confirmed that the challenge either escalated or solidified their path towards Healthcare AI — the highest "intent-to-pursue" signal we have ever recorded for a student competition at UPM.
Winners

Champions & the people who built them up

Top 5 of IDSC 2026 (Grand Final — 14 April 2026)

PlaceTeamInstitutionPrize (RM)
1st (Champion)BioinformersUniversiti Malaya (UM)500
2ndGAMA-BCIUniversitas Gadjah Mada (UGM)300
3rdMultiUUM · USM · UPM200
4thBANG SADAUniversitas Brawijaya (UB)100
5thVentriculearn-11UKM · Universitas Bengkulu100

Full Top 10 Finalist Roster

TeamMembersInstitution
GlaucoNetKhairunnisa Maharani, Rahmah Gustriana DekaInstitut Teknologi Sumatera (ITERA)
Ventriculearn-11Ahmad Hathim, Anis Syauqina, Elvira Yunita, Siti Nor AmiraUKM · Universitas Bengkulu
BioinformersKumanan, Wisely, Khoo Li Ying, KavitaUniversiti Malaya (UM)
Team 108Lee Pei En, Ng Zhi Ying, Sim Yu Yin, Yah Tian LingUniversiti Malaya (UM)
VZIzzar Sully Nashrudin, Risma MuslimahUIN Maulana Malik Ibrahim Malang
MultiUSaw Yong Quan, Tan Shan Qi, Hoe Zhi Wan, Hong Tze LoonUM · USM · UPM
abjritAli Zainal Abidin, Najwa Fadhilah, Maulida Rahmi, Rafidah KhoirunnisaInstitut Teknologi Sepuluh Nopember (ITS)
GAMA-BCIFarichaturrifqi Aryanitasari, Muhana Fawwazy Ilyas, Hamzah Arman Husni, Aulia Gita PratiwiUniversitas Gadjah Mada (UGM)
BANG SADANathanael, Adinda, Bonfillo, KeenanUniversitas Brawijaya (UB)
FD2Ahmad Naim, Muhammad Akmal, Ahmad Mukhlis, Muhammad Shahril AdanUniversiti Malaya (UM)

Workshop leaders — the people who set the tone (14 Mar 2026, International Mathematics Day)

Data Science & Machine Learning

Assoc. Prof. Ts. Dr Shukor Sanim Mohd Fauzi — UiTM Perlis. Advisory Group Member, National AI Office (NAIO), Ministry of Digital Malaysia. Former Rector of UiTM Perlis; Chairman of the Northern Region Big Data Research Group. Secured over RM 4.1 million in research funds with 161 scientific publications. Applies custom CNN pipelines to deep predictive environments (e.g. wildfire forecasting). Workshop helpfulness: 4.46 / 5.

Presentation Skills

Datin Dr Ghazila Ghazi — Senior Lecturer, Faculty of Modern Languages & Communication, UPM. Specializes in political, public oration & law enforcement communication. Co-developed digital policing frameworks with the Royal Malaysia Police (VSP platform). Former broadcast journalist on TV3's 999 and Majalah 3. Taught teams how to translate algorithmic accuracy into a story a Director will remember.

Stage 2 — expert judging panel

JudgeAffiliationAssigned domain
Prof Dr Nur ChamidahUNAIR · Professor of StatisticsC1 — Mathematical & Algorithmic Rigor (25 pts)
Assoc Prof Dr Shukor SanimUiTM · Assoc. Prof Software EngineeringC2 — Model Performance (20 pts) · C4 — Innovation (15 pts)
Dr NasheefHospital Dalat · Director (Sarawak)C3 — Problem Definition & Healthcare Relevance (15 pts)
Dr NaimHospital Tengku Permaisuri Norashikin · Deputy DirectorC5 — Clinical Interpretability & Safety (10 pts)
Dr GhazilaUPM · Senior Lecturer, CommunicationC6 — "Hope" Impact (10 pts) · C7 — Presentation Quality (5 pts)

The full Stage 1 preliminary panel (33 judges)

#ExpertInstitution#ExpertInstitution
1Ahmad Hakiim JamaluddinUPM18Nariza Wanti Wulan SariUNMUL
2Nur Ezlin ZamriUPM19Shahirah Abu BakarUTM
3Nur Syahirah WahidUPM20Rusya Iryanti YahayaUUM
4Siti Maghfirotul Ulyah, Ph.DUNAIR21Nor Ain Azeany Mohd NasirUPNM
5Toha SaifudinUNAIR22Hafizah Farhah Saipan @ SaipolUTM
6M. Fariz Fadillah MardiantoUNAIR23Samsul Ariffin Abdul KarimUUM
7Marisa Rifada, M.SiUNAIR24Ts. Dr Latifah Abd LatibUPM
8Hani Syahida ZulkafliUPM25Saadi Ahmad KamaruddinUUM
9Syaiful AnamUB26Syarifah Zyurina NordinUTM
10Avin MaulanaUB27Mohamad Huzaifah DzubaidiUUM
11Hilmi Aziz BukhoriUB28Nurhazimah NazmiUTM
12Nur Silviyah Rahmi, M.StatUB29Siti Nur Ainsyah GhaniUUM
13Muhammad Aslam Mohd SafariUPM30Nur Syahirah IbrahimUUM
14Meirinda FauziyahUNMUL31Dita AmeliaUNAIR
15Andrea Tri Rian DaniUNMUL32Mohd Shafie MustafaUPM
16Memi Nor HayatiUNMUL33Farid Zamani Che RoseUPM
17Regita Putri PermataTelkom Univ.———
Rubric

The rubric we used — published on Day 1, kept till Day 60

Total available: 100 pts. The categories below were publicly visible from the registration page, so teams could strategise around clinical safety and "hope" impact (not just raw accuracy).

Stage 1 & Stage 2 unified evaluation rubric
CriterionWeightage
Mathematical / Algorithmic Rigor25%
Model Performance & Validation20%
Problem Definition & Healthcare Relevance15%
Innovation & Creativity15%
Clinical Interpretability & Safety10%
"Hope" Impact & Practical Implications10%
Presentation Quality & Clarity5%
Total100%

Tie-break protocol

  1. Higher score on C1 — Math/Algorithmic Rigor.
  2. Then higher score on C3 — Problem Definition & Clinical Relevance.
  3. Then higher score on C5 — Clinical Interpretability & Safety.
  4. Final: rapid-fire deliberation panel + majority vote by the 5 judges.
Why this matters. Three of the seven categories explicitly reward clinical literacy. Treating those as "soft criteria" is how competitions promote clinically useless models. Don't do it.
Judges' Notes

What the judges wrote on the winners

Cross-referenced from the master evaluation board spreadsheet. These four quotes are the checkpoints the next-generation rubric should keep rewarding.

"Delivers an exceptionally rigorous approach to Brain-Computer Interface (BCI) technology for ALS patients. The team successfully addressed a major clinical bottleneck: the need for daily BCI recalibration. Brilliant presentation, and the models are highly feasible for direct clinical usage."— On Bioinformers (1st place) · BCI track
"Presents a highly sophisticated, multimodal approach to ALS communication by fusing EEG and eye-tracking data. Their standout triumph is the design of an uncertainty-aware safety gate and their rigorous zero-shot evaluation on actual ALS patients — beautifully bridges algorithmic accuracy and clinical trust."— On GAMA-BCI (2nd place) · BCI track
"The team excels in translating physiological realities into smart data science decisions, brilliantly targeted the exact metric primary care physicians care about. By achieving a 94.5% NPV, the model safely cleared 69 healthy patients — successfully achieving their goal of providing safe 'discharge confidence' to front-liner Medical Officers."— On Ventriculearn-11 (5th place) · SPECT track
"The team implemented a 'Threshold Innovation' strategy, explicitly defining their problem as creating a 'zero-miss' safety tool where False Negatives are clinically unacceptable. 100% recall on an independent test set with patient-level validation directly addresses the highest clinical safety concern."— On Team 108 · ECG track
"Their problem definition is exquisitely structured around the physiological reality that patients are the core unit of diagnosis, not individual images. Moved from image-level to patient-level decision-making — reducing false negatives from 32 down to 2."— On BANG SADA · Glaucoma track
"Grad-CAM explicitly demonstrated with anatomically correct activation on the optic disc for positive cases, and appropriately diffuse activation for negatives. The team articulated XAI in clinically meaningful terms — 'auditable, anatomically grounded AI'."— On GlaucoNet · Glaucoma track
For INNOVATHON 2026

Six quick lessons I will steal for INNOVATHON 2026

If you only have takeaways from this talk, take these six — and apply them before lunch.

1. Reward publicly and on the same day

RM 500 / 300 / 200 plus MPC case-study slot, certificate, trophy. Make the win feel like a national exam result — families watch the livestream.

2. Publish the rubric early

Eight criteria: problem, creativity, user-friendliness, deployment, validation, outcome, scalability, sustainability. No surprises on stage day.

3. Use legitimate judges

Mix clinicians, statisticians, journalists, policy people. Domain beats glamour every time.

4. Mentor in public, online, three weeks

Fixes 80% of weak submissions at zero extra cost. Just do it.

5. Tell stories, not scores

Judges love a "100% recall" headline; nurses love a "52-second turnaround" headline. Lead with the turnaround.

6. Celebrate the closing ceremony like a national award

Floor managers, livestream, photos, free lunch, hashtags. Treat the closing like an MPC ceremony, not a meeting room.

Bonus tip for Malaysian hospital committees. Build the rubric with the Senior Consultants in the room. Same coffee, same table. Avoid the meeting-after-the-meeting culture that quietly kills innovation budgets by August.
For INNOVATHON competitors

To the teams competing today — read this twice

If you are pitching in this hall this week, here is what the IDSC 2026 evaluation board actually rewarded — straight from the judges' comments and the participant feedback log.

What to bring on stage

  1. One specific patient journey. Not "cardiology" — "the ED triage queue at HTPN Kajang on a Saturday night."
  2. One specific clinical safety number. Recall, NPV, false-negative cost. Not accuracy.
  3. One specific clinician workflow change. What does the nurse stop doing? When does the consultant get paged less?
  4. One specific deployment plan. Which station? Which kiosk? Which phone?
  5. One story for the Director's mother. Don't let the rubric eat your humanity.

What kills the pitch

  • Models with no patient-level validation (only image-level F1).
  • "We used a transformer" with no clinical-safety narrative.
  • Datasets patched together from open sources without consent trail.
  • Promising deployment "at Hospital X" without a single line of clinical engagement.
  • Ignoring the NPV — NPV is what makes the Medical Officer sleep at night.
What participants told us mattered most. "Mathematics was not only about formulas, but about transforming medical data into meaningful insights. We learned that healthcare technology must balance accuracy, fairness, and responsibility. A model with high accuracy is still not enough if it cannot be trusted or applied fairly to different patients. Errors can affect real lives." — Team DMAZA.
Another participant's framing. "A single false negative in Brugada Syndrome could mean losing the opportunity to prevent sudden cardiac death. We prioritized patient safety and clinical utility over maximizing performance metrics. For our team, 'hope' means using AI to give doctors better tools and a greater chance to save lives." — Team Muadz.
For Malaysian clinicians

To the clinicians in the room — the Monday-morning version

You don't need a national AI strategy to start. You need a small Team A that finishes one quick-win by next month, and a Team B that promises to copy it.

Build the rubric on a napkin

Sit with two consultants, one Medical Officer, one nurse, one IT lead. Write down what "good enough to use tomorrow morning" looks like. That is your rubric.

Find three pain points — pick one

Don't run a hospital-wide brainstorm. Run a 30-minute round-table with five senior people. Pick the one pain point that costs the most in minutes, not in theory.

Reward in front of the DG

The most under-used reward in Malaysian healthcare is visibility. When the DG, the State Health Director, and the press are in the room, the team will remember it for ten years.

A 90-day quick-win ladder for a Malaysian hospital

Days 1–14 — pick the pain

Round-table with senior clinicians; lock one pain point with measurable minutes-saved per shift. Locks: rubric, dataset, success metric.

Days 15–45 — small team, fast build

One data scientist, one domain expert, one UX. Build a shadow-mode tool — it predicts, but the clinician still decides.

Days 46–70 — pilot on one ward / one clinic

Track three numbers: minutes saved, recall, and clinician trust (yes/no). Publish the dashboard internally, in plain Malay/English.

Days 71–90 — celebrate publicly, then scale

Run a tiny mini-innovathon with the next ward. Award visibility + a small certificate + a coffee with the HOD. Recruit the next Team A.

The compounding effect. One quick win in 90 days = four quick wins in 12 months = a department that actually adopts AI on its own terms. The NHS used this pattern for decades. Malaysian hospitals can too — without massive capex, without a vendor lock-in, without a 5-year roadmap.
Behind the Scenes

How we organised IDSC 2026 — the structural framework

Organising institutions

  • Main: Dept. of Mathematics & Statistics, Universiti Putra Malaysia (UPM)
  • Co: Dept. of Mathematics, Universitas Airlangga (UNAIR), Indonesia
  • Co: Dept. of Mathematics, Universitas Mulawarman (UNMUL), Indonesia
  • Co: Dept. of Mathematics, Universitas Brawijaya (UB), Indonesia
  • Partner: PERSAMA

Grand-Final command centre

Department of Mathematics & Statistics Meeting Room, UPM · Broadcast and jury managed on-site, all teams joined live from five countries.

Advisory board & executive committee

BoardOfficerAffiliation
Academic AdvisorsProf. Dr Norihan Md ArifinUPM
Prof. Dr Leong Wah JuneUPM
Industry AdvisorsDr Izzun Nasheef M. HasmuriDirector · Hospital Dalat, Sarawak
Dr Naim Abdul MalekDeputy Director · Hosp. Tengku Permaisuri Norashikin
Core LeadershipDr Ahmad Hakiim JamaluddinOrganising Chair (UPM)
Dr Farid Zamani Che RoseCo-Organising Chair (UPM)
Dr Mohd Shafie MustafaCo-Organising Chair (UPM)
Secretariat ExecsDr Nur Syahirah WahidSecretariat Lead (UPM)
Ms Nur Aqilah Mohd NoordinSecretariat Lead (UPM)
Finance ExecsDr Nur Ezlin ZamriFinance Lead (UPM)
Dr Hani Syahida ZulkafliFinance Lead (UPM)
Judging Exec BoardDr Muhammad Aslam Mohd SafariJudging Lead (UPM)
Dr M. Fariz Fadillah Mardianto, M.SiUNAIR Representative
Dr (Cand) Andrea Tri Rian DaniUNMUL Representative
Assoc. Prof. Dr Syaiful AnamUB Representative

Student organising taskforce

Technical & Submissions

Aleeya Natasya Azahar · Nuratika Anuar

Secretariat & Participants

Sametha Sivalingam · Nurul Ayuni Omar · Francis Lim Beng Cong

Publicity & Design

Mohamad Shahrul Ikram · Thurgashini Gunalan · Nurul Edlina Shazwan

Finance & Admin

Puviniyaraj S.Rajendran · Cheah Hui Yin

National Impact

The "Hope" algorithm — national press & career outcomes

The event's impact was officially captured by Dewan Kosmik by Dewan Bahasa dan Pustaka (DBP), Malaysia's national language and publishing agency. That is the kind of visibility the Reward Culture Framework calls for in pillar #4.

Career-trajectory signal

Response% of respondents
Significantly increased — now plan to pursue Healthcare AI42%
Moderately increased29%
Confirmed existing interest25%
No change4%

A few sample cross-university teams (from the official registry)

Team#Leader InstitutionLevelAll institutions in the alliance
1Braincore IndonesiaPGBraincore Indonesia · Universitas Brawijaya (UB)
13Universiti Kebangsaan MalaysiaPGUniversitas Bengkulu · UKM
85UKMUGAnna University (India) · UKM
112Universitas Brawijaya (UB)UGUB · Universitas Udayana (UNUD)
140UIN MalangUGDushanbe Innovation Institute (Tajikistan) · UIN Malang