Software · AI · Systems

Shachindran
Veerangan

I build software for messy, real-world operations.

B.IT (Hons) · Universiti Teknologi PETRONAS · 2027Malaysia · Ankara · Bangkok
Selected work

Software for the parts of work
that are still too manual.

01 / 06

GASOS

Vertical SaaS · LPG distribution

One connected platform for LPG distribution.

Designed and engineered for the daily realities of LPG businesses, GasOS connects orders, deliveries, cylinder inventory, invoicing, payments and Buku Penyata reporting. The platform is expanding into native accounting, expenses, fleet operations and business management within one integrated product.

Read the full case study
GasOS platformFrom order to business-day close
GasOS / 01
01Order & dispatchCoordinate the work
02Cylinder custodyTrace physical movement
03Invoice & receivableLink delivery to what is owed
04Payments & cashRecord collections and handovers
05Buku PenyataBuild source-linked reporting evidence
06Daily closeReview differences and exceptions
GasOS Office + DeliveryConnected operations, built for LPG

GASOS

Making the physical and financial sides of LPG distribution agree.

01Order & dispatchCoordinate the work
02Cylinder custodyTrace physical movement
03Invoice & receivableLink delivery to what is owed
04Payments & cashRecord collections and handovers
05Buku PenyataBuild source-linked reporting evidence
06Daily closeReview differences and exceptions

GasOS brings orders, deliveries, cylinder inventory, invoices, payments, cash handovers and Buku Penyata into one connected workspace. Development is expanding the platform with accounting, expenses, fleet management and additional business operations.

Read the full case study
Bangkok · 2026

AI / NLP R&D · Omniscien Technologies

I worked on document intelligence: turning messy files into structured information a system could actually use.

I built and tested parts of a document ingestion and enrichment pipeline across PDFs, DOCX files, email, scans and less common formats. The work covered extraction, table handling, metadata, OCR and correction, plus the async services around the pipeline.

FastAPICeleryRedisDockerKubernetes
PDFReports
DOCXDocuments
EMLEmail
IMGScans
ARCArchives
Document intelligenceStructured output
texttablesmetadatadetails
01Ingest

Bring very different document types into one asynchronous processing path without treating every file as a special case.

02Extract + enrich

Pull out useful text, tables and metadata, then improve difficult output with OCR and correction where needed.

03Run it as a system

Work across APIs, background jobs, Redis, containers and the Kubernetes deployment context around the pipeline.

Bangkok · 2026

Document intelligence

AI / NLP R&D · Omniscien Technologies

I worked on turning messy documents into structured information a system could actually use.

PDFReportsDOCXDocumentsEMLEmailIMGScansARCArchives
FastAPICeleryRedisDockerKubernetes
01Ingest

Different formats, one asynchronous processing path.

02Extract + enrich

Text, tables, metadata, OCR and correction.

03Run it as a system

APIs, background jobs, containers and Kubernetes context.

Three smaller builds. Three different problems.

Predictive modeling, time-series forecasting and a motion-heavy web experience — useful snapshots of how I approach different kinds of work.

03.1Industrial AI

Predictive
Maintenance

A remaining-useful-life model built on NASA's C-MAPSS turbofan dataset, using sensor degradation data to estimate how many cycles an engine has left before failure.

PythonScikit-learnRandom Forest
View repository
NASA C-MAPSS / FD001Remaining useful life
Predicted versus actual remaining useful life results from the predictive maintenance project
Input21sensors
Context03operating settings
TargetRULcycles
ModelRFrandom forest
Sensor degradationRegressionPredict before failure
03.2Time series

Smart Energy
Forecaster

An hourly electricity-demand forecaster built around AEP data, with daily, weekly and seasonal patterns surfaced instead of hiding the forecast behind a single number.

PythonProphetAEP data
View repository
AEP / hourly demand24H
evening peak18:00–20:00
DailyWeeklyYearly seasonality

Malaysia → Ankara → Bangkok.

Three places shaped three different parts of how I work: UTP gave me the foundation, Ankara made me adaptable, and Bangkok taught me to own real engineering problems.

Portrait at Universiti Teknologi PETRONAS during an award ceremony04.1
Malaysia2023 — present
Universiti Teknologi PETRONAS logoUniversiti Teknologi PETRONAS

UTP

I’m completing a Bachelor of Information Technology (Hons) at UTP. The degree gave me a broad software foundation across data science, databases, object-oriented programming and software engineering. Outside class, leadership roles taught me how to coordinate people, make decisions and deliver when the work was bigger than one person.

Engineering foundationLeadershipOwnership
Bachelor of Information Technology (Hons) · YUTP Education Grant · Dean’s List
Portrait on a Bilkent University bench in Ankara04.2
Türkiye2025
Bilkent University logoBilkent University

Ankara

I spent a semester at Bilkent University as an exchange student in Computer Information Systems and Technologies. It was the first time I had to rebuild my routine in a different academic and cultural environment, and I learned to adapt quickly, communicate across backgrounds and become comfortable starting from zero. It made an international career feel realistic rather than distant.

AdaptabilityCross-cultural communicationIndependence
Exchange Student · Computer Information Systems and Technologies · Bilkent University
Portrait at a technology conference in Bangkok04.3
Thailand2026
OTOmniscien Technologies

Bangkok

Bangkok was where software stopped feeling like coursework. In AI/NLP R&D at Omniscien, I worked on document intelligence across heterogeneous file ingestion, extraction and enrichment, OCR paths, asynchronous workers and containerized environments. I learned to benchmark before choosing, trace problems through a system, test the awkward cases and keep moving when the answer was not obvious.

Document intelligenceSystems thinkingEngineering under uncertainty
AI / NLP R&D Intern · Document Intelligence · Omniscien Technologies

Most things I build start with a broken workflow.

I’m an Information Technology undergraduate at Universiti Teknologi PETRONAS working across software systems and applied AI. I’m drawn to problems where the real process is messier than the software around it.

That is what pulled me into GASOS and document intelligence. I’m most interested in systems that have to deal with imperfect inputs, real operational constraints and people who need the software to work — not just demo well.

EducationBachelor of Information Technology (Hons)

Universiti Teknologi PETRONAS · Expected 2027

ExchangeBilkent University

Ankara · 2025

ExperienceAI / NLP R&D

Document intelligence · Bangkok · 2026

Current directionSoftware systems · Applied AI

Document intelligence · Multimodal systems

AcademicYUTP Education Grant · Dean’s List
LeadershipHoliRun · Assistant Project Director

250+ participants · RM29,390 programme budget

Working on something difficult?

I’m open to engineering roles, postgraduate opportunities and a small number of collaborations.