Calvin
Joy
Tarigan
AI/ML · Software Engineer
Computer Science student at Universitas Indonesia focused on building and implementing machine learning systems — from custom neural networks and RL agents to production-ready full-stack platforms.
Selected Projects
See all →Support Agent
Customer support agent with native tool-calling, RAG with ChromaDB, and 85.7% tool-call success rate.
Ops Agent
Multi-agent IT ticket resolution with LangGraph, HITL approval, and 90% routing accuracy.
Vacuum Cleaning Robots MDP
Modular RL experimentation framework benchmarking 6 algorithms on a 7x7-grid vacuum environment.
Projects
A collection of AI/ML and software engineering work.
Support Agent
Built a customer support agent from scratch in Python implementing a native tool-calling loop without agentic frameworks. Integrated RAG with ChromaDB and Voyage AI embeddings, engineered 4 tool functions with guardrails, and achieved 85.7% tool-call success rate on a 20-case golden evaluation dataset. Implemented circuit-breaker failover to a secondary LLM provider and CI/CD with 13 automated pytest cases via GitHub Actions, containerized with Docker and deployed on Hugging Face Spaces.
Ops Agent
Designed a multi-agent orchestration system using LangGraph with a supervisor/router pattern coordinating Technical, Billing, and Security agents. Implemented human-in-the-loop approval for sensitive operations and RAG-backed runbook retrieval via ChromaDB with persistent checkpointing. Achieved 90% routing accuracy on a 30-ticket golden evaluation dataset with 16 automated tests.
IDX Multi-Asset Forecasting
Built an end-to-end deep learning pipeline to forecast 15-minute future log returns for 100 active Indonesian stocks using intraday data from 787 IDX tickers. Implemented and compared GRU, LSTM, and TCN architectures; the best-performing 2-layer GRU achieved validation RMSE of 0.006184. Conducted systematic ablation studies revealing architectural design trade-offs.
Vacuum Cleaning Robots MDP
Built a modular Gym-based experimentation framework benchmarking 6 RL algorithms on a 7x7-grid vacuum-cleaning environment. PPO emerged as the best policy with 89.6% clean-cell ratio and 71.7% task success rate. Automated experiment runs and evaluation logging for reproducible analysis, documented in a structured LaTeX technical report.
NusaWallet
Semifinalist — Hackathon Digdaya BI 2026. Built a multi-currency mobile application supporting the digitalization of export services for freelancers and digital MSMEs in Indonesia. Backend built with Django; mobile client built with React Native (Expo), integrated with live foreign-exchange rates via the Frankfurter API.
Smart Garden
Built a soil-moisture monitoring system on the AVR XMEGA A3BU microcontroller platform. Extended the system with an ESP32-based soil moisture and automated lighting controller.
Trashmate
“Inovasi Digital untuk Pengelolaan Sampah yang Terintegrasi” — a full UI/UX research case study for an integrated digital waste-management solution. Carried the project end-to-end through the UX research lifecycle: user research, persona development, prototyping, and usability evaluation.
SiSidang — University Court-Scheduling System
Worked in a 7-member Scrum team over 4 sprints to deliver stakeholder-driven features and bug fixes for a production web application. Maintained and refactored a legacy Django/PostgreSQL backend paired with a Next.js frontend across 4 repositories. Used Jira, GitLab CI/CD, Docker, SonarQube, New Relic, and OWASP ZAP for project management, deployment automation, and security assurance.
Rewear+
Silver Medal — Ignite Future Fest 2026 National Essay & Business Plan Competition. Authored a business plan proposal on technology-driven digital transformation and built a supporting prototype for Rewear+, a fashion resale marketplace concept, using Next.js and TypeScript.
Little Margo Catapult
Solved a digital-presence problem for Little Margo, a local UMKM, by building a company profile website using Next.js and TypeScript deployed on Vercel. Collaborated in a 2-person team where both developers worked full-stack.
EventSphere
Built backend with Java Spring Boot and a responsive frontend with Next.js backed by PostgreSQL. Designed and integrated RESTful APIs for comprehensive event creation, management, and participation features.
HealthMate
Collaborated on a comprehensive healthcare management web application across 3 repositories backed by PostgreSQL. Core monolith built with Django and HTML/CSS frontend with 2 additional microservices for OTP and phone verification. Contributed to containerized deployment using Kubernetes and implemented advanced security measures for sensitive medical data.
SIZOPI — Zoo Management System
Developed a Django-based monolith backend with full CRUD operations and an HTML/CSS frontend, contributing to PostgreSQL schema design. Created efficient data models for tracking animal care, feeding schedules, and visitor information.
Dinepasar
Eksplorasi cita rasa khas kuliner Denpasar — built a responsive web and mobile platform for restaurant discovery featuring menus, ratings, and user reviews across 2 repositories. Web client built with Django and HTML/CSS; mobile client built with Flutter (Dart), both backed by a shared Django backend.
Experience
Teaching, mentoring, and organizational roles.
Teaching Assistant
Delivered course material on software security concepts and supported students through lab sessions and assignments.
Coordinator of Teaching Assistant
Led a team of TAs coordinating weekly sessions, grading, and student support. Strengthened ability to communicate mathematical reasoning, directly supporting analytical skills in ML coursework.
Programming Fundamentals Mentor
Mentored students with no prior coding background on programming fundamentals. Developed teaching materials and structured exercises; communicated complex technical topics to non-technical audiences.
Staff — Sports Department
Organised and coordinated faculty-wide sports activities and competitions. Developed teamwork and event management capabilities.
Universitas Indonesia
Supervised Machine Learning: Regression & Classification
Custom Models, Layers, and Loss Functions with TensorFlow
Technical Skills
Get in Touch
Open to internships, collaborations, and research opportunities.
- Internships & Part-time roles
- ML/AI Research Collaboration
- Open Source Contributions
- Data Science Freelance Projects