ARTIFICIALINTELLIGENCEENGINEER
AI Engineer specializing in Generative AI, Computer Vision, and predictive modeling. I build and deploy end-to-end AI-powered applications, from RAG-based chatbots and OCR pipelines to deep learning research systems.
WHAT I BRING TO THE TABLE
Automate Document Workflows
I build RAG chatbots and OCR pipelines that turn hours of manual document processing into seconds of automated intelligence.
Ship Computer Vision Systems
From drowsiness detection to waste classification, I deliver vision models with measurable, validated accuracy.
Turn Data Into Decisions
I design dashboards and predictive models that replace gut feelings with data-driven insights, reducing manual reporting by 40%.
CONFIDANT PROFILE
Adji Dharmawan Indrianto
AI Engineer
Information Engineering graduate from Universitas Gadjah Mada, specializing in Generative AI, Computer Vision, and predictive modeling. I architect and deploy production AI systems that solve real business problems, from RAG-based document intelligence chatbots and automated OCR pipelines to deep learning research systems such as transformer-based drowsiness detection. My work spans the full lifecycle: research and experimentation through to end-to-end MLOps deployment with Docker, FastAPI, and cloud platforms. Along the way I've won multiple national machine learning and data science competitions, including 1st place at Alphathon 2026. I've built AI systems at PT PLN and Viktori Askara Teknologi Indonesia, and I now work as an ML Engineer at MAP Active, driven by the goal of bridging cutting-edge research with tangible, real-world impact.
Universitas Gadjah Mada
B.Eng. Information Engineering • GPA: 3.69/4.00
Coursework: AI, Data Engineering, Cloud Computing, Algorithms
◆ AI ARSENAL
PALACE INFILTRATION LOG
Data and AI Engineer
Building and deploying production machine learning and LLM-powered systems on enterprise cloud AI platforms, including Microsoft Azure AI Foundry and AWS Bedrock.
AI Engineer Intern
Engineered a Decision Support System (DSS) consolidating multi-domain datasets (sales, subsidies, compensations, and voltage drops), reducing executive reporting time by an estimated 40%. Architected and integrated a Generative AI-powered chatbot into the DSS dashboard to answer complex ad-hoc analytical queries, cutting manual data-gathering requests by an estimated 30%.
Data Science Mentor
Mentored student teams for the BDC Satria Data 2026 competition, designing and teaching a 3-class waste image classification pipeline that reached an F1-Macro of ~0.995 under nested cross-validation. The pipeline guards against data leakage with pHash deduplication and StratifiedGroupKFold, and ensembles frozen Vision Transformers (PE-Core Gigantic and SigLIP2) with custom PyTorch MLP heads.
AI Engineer Intern
Architected and deployed an end-to-end Document Intelligence Chatbot using RAG with optimized LLM inference, serving ~150 daily internal queries and reducing document retrieval time by 35%. Implemented a double-layer PII sanitization pipeline that cleared internal privacy review with zero PII leakage findings, and engineered an automated OCR pipeline processing 500+ financial documents/month at 92% accuracy, accelerating credit assessment by 40%.
Data Analyst Intern
Built dynamic BI dashboards in Power BI and Metabase, integrating Google Analytics to monitor web traffic and financial KPIs, cutting manual reporting time by 40%. Processed 10,000+ rows of financial data with Advanced Excel (Power Query, Power Pivot) and SQL, and optimized production Metabase queries to reduce execution time by 60%.
SELECTED TARGETS
Water Quality Anomaly Detection
Unsupervised machine learning framework detecting river water contextual anomalies using Causal Z-Score feature engineering and a heterogeneous voting ensemble.
Driver Drowsiness Detection
High-precision drowsiness detection using the VideoMAE V2-Huge transformer with LoRA fine-tuning. Reached a 0.8795 F1-Score while training less than 2% of the parameters.
Smart Fridge Manager (Anti-Basi)
AI-powered food detection using the Gemini Vision API with Firebase real-time inventory tracking and automated expiry notifications via Cloud Messaging.
FIELD NOTES
The Blueprint of LLM Fine-Tuning
A guide to modern LLM fine-tuning across two paradigms: the compute side (Full Fine-Tuning vs. PEFT methods like LoRA and QLoRA) and the objective side (SFT, RLHF, and DPO alignment), balancing memory efficiency with human alignment.
The Evolution of RAG: From Naive to GraphRAG
How RAG architectures evolved from Naive to GraphRAG, and how to choose the right pipeline to keep LLM answers grounded in your own data.
Awakening the Giant: How We Tamed VideoMAEv2-Huge
A field report on fine-tuning a billion-parameter video transformer on constrained hardware: LoRA adapters, frozen backbones, and the memory tricks that made VideoMAEv2-Huge trainable.
Beyond Traditional ML: Why and When We Actually Need Deep Learning
Breaking down the bottlenecks of traditional ML and identifying the exact scenarios where architectures like CNNs and Transformers become a necessity.
SKILLS ACQUIRED
BATTLE RESULTS
1st Place • Alphathon 2026
Data Science Indonesia & WorldQuant Brain • 2026
Won the national alphathon by building alphas (mathematical models that predict stock price movements) on WorldQuant's BRAIN platform, using historical market data and predefined operators to construct and backtest quantitative equity strategies.
1st Place • ML Competition Data Slayer 3.0
Telkom University Purwokerto • 2025
First place for developing a real-time driver drowsiness detection model, classifying signs of fatigue from live input to support road-safety applications.
1st Place • Data Science Competition
TETI Programming Week, UGM • 2025
Won the competition with a network-traffic classification model that detects whether traffic belongs to a botnet, supporting cybersecurity threat detection.
3rd Place • Data Science Competition
Informatics Festival 2026, Universitas PGRI Semarang • 2026
Third place for a predictive model that estimates an MSME's (UMKM) likelihood of success based on key business factors.
Finalist • Student Digital Innovation
Kemendikbudristek RI • 2025
National finalist for a VR-based disaster-preparedness learning app, enhanced with an integrated chatbot to guide users through natural-disaster education.
LET'S STEAL
SOME HEARTS
Open to AI/ML engineering roles, freelance projects, or collaborative research.
ML Engineer at MAP Active. Open to select freelance AI/ML projects and research collaborations.
Jakarta, Indonesia