Updated 16 November 2026
Allan Emmanuel B. UMALI, ECE, ECT
Applied AI & Computer Vision · Open to graduate research opportunities in Japan
meet.allanemmanuelumali.com | allanemmanuelumali@gmail.com | linkedin.com/in/emsumali | Manila, Philippines
Research Interests
Computer vision for real-world systems: self-auditing traffic analytics, retail behavior sensing, and edge deployment. I also work on reliable LLM agents, which means evaluation design, guardrails, human-in-the-loop workflows, and cost-aware model routing, along with data-efficient annotation, synthetic data, and MLOps reproducibility.
Publications
- Umali, A.E.B., et al. "GATE: Graph-Admissible Transition Evaluation for Self-Auditing Turning-Movement Counting from Monocular Traffic Cameras," in 2026 IEEE International Conference on Future Machine Learning and Data Science (FMLDS), Kobe, Japan, Nov 2026. Accepted, to appear in the proceedings. First author and presenter.
- Jose, J.A.C.; Bertumen, C.J.B.; Roque, M.T.C.; Umali, A.E.B.; Villanueva, J.C.T.; TanAi, R.J.; Sybingco, E.; San Juan, J.; Gonzales, E.C. "Smart Shelf System for Customer Behavior Tracking in Supermarkets," Sensors (MDPI), 24(2), 367, 2024. doi:10.3390/s24020367. Contributed the weight-sensor system, the off-the-shelf vision system, and customer-product association in the integration system.
Education
De La Salle UniversityManila, Philippines
M.Sc. in Data Science (in progress; coursework partway complete)Sep 2025 – Present
- Advanced machine learning, statistical modeling, and big-data analytics.
De La Salle UniversityManila, Philippines
B.Sc. Electronics and Communications Engineering, minor in Applications DevelopmentSep 2019 – Dec 2023
- Thesis: "Smart Shelf System for Customer Behavior Tracking in Supermarkets." Gold Undergraduate Thesis Nominee; the work was later published in Sensors (2024).
- Licensed ECE and ECT (October 2024), general average 85.70%, among the top ratings of DLSU takers.
- Jose Rizal Honors; Dean's Lister six times; Philippine National Champion, IECEP Interscholastic Quiz Show 2022.
Research Experience
De La Salle University, Intelligent Systems LaboratoryManila, Philippines
Project Technical Associate IIJan 2025 – Jan 2026
- Led an end-to-end AI vehicle-counting and traffic-policy system: real-time detection and counting across camera angles, plus traffic forecasting for holidays, special events, and ordinary days.
- Co-supervised research teams totaling 8 members: undergraduate researchers on API development, automation, and edge-case analysis, and interns on CVAT annotation quality. Supervised a professional transitioning into AI, including a comparative study of auto-annotation tools.
De La Salle University, Intelligent Systems LaboratoryManila, Philippines
Research Assistant (full-time Dec 2023 – May 2024; part-time 2020 – May 2023)2020 – May 2024
- Supported the smart-shelf thesis research through experiment design, data collection, system integration, and validation.
- Contributed to active-learning research that reduces annotation effort for retail visual analytics, and to cloud- and network-based automated video annotation.
- Led component research and power modeling for an autonomous mobile robot with SLAM and speech recognition: NVIDIA Jetson Xavier NX, Intel RealSense, and LiDAR across 17 suppliers.
Applied AI Experience
KDCI (Key Discovery Consulting, Inc.)Pasig City, Philippines
AI Specialist (AI Team Lead)Jan 2026 – Present
- Founding engineer of the company's AI team. I take agentic-AI systems to production with evaluation harnesses and guardrails.
- Evaluation design. A 250-sample model A/B test raised verified evidence from 86.9% to 99.0% at ~5× lower cost when classifying a ~1,600-provider AI-vendor directory (CoreIndex).
- Human-in-the-loop customer-service AI. Graded draft accuracy rose from 36% in the first graded week to 97% in the latest, sustaining 90%+ monthly accuracy, with the last two graded weeks at 95.5% (above the 95% target). The error rate fell from 71% to 8%. Every reply is human-approved.
- Voice AI. Built Elaine, the company's inbound voice agent on kdci.co. It answers from a RAG knowledge base, and quoted prices come only from a pricing tool, never from the model.
- Enablement. Top-rated trainer (4.98/5, NPS +90) in the company's AI Foundations pilot, and author of its AI data-governance framework.
Supervaise, Inc.Manila, Philippines
Artificial Intelligence EngineerJan 2025 – Jan 2026
- License-plate recognition. Re-architected the pipeline from 108.6 s to 11.2 s runtime (−90%). Contributed to a system reaching 90.61% detection mAP and 90.08% recognition accuracy, deployed on NVIDIA Jetson.
- MLOps. Built retraining automation (TAO Toolkit, YOLOv4-tiny, pruning, FP16/INT8 export) with experiment tracking, dataset fingerprinting, and reproducibility packages. Found and fixed a validation-set data-leakage flaw in the benchmark pipeline.
- Annotation and synthetic data. Pre-annotation with T-Rex2 and SAM halved annotation time, and I explored ComfyUI / Stable Diffusion domain randomization to adapt models to new camera sites.
Teaching & AI Education
- Teaching Assistant, DLSU (Jul 2022 – Jan 2026). Supported 500+ engineering students in Programming Fundamentals (C, MATLAB) and the Intelligent Systems Laboratory (k-NN, gradient descent, neural networks).
- NVIDIA AI Philippines (Oct 2022 – Jan 2026). Founding team member and technical lead of a 2,000+ member community. Organized and co-taught NVIDIA Deep Learning Institute workshops, including a hybrid workshop with 252 participants from 17+ universities.
- Government and enterprise AI workshops (2025). Developed curriculum and taught at regional-government and enterprise AI-enablement workshops on prompt engineering, RAG research workflows, and AI assistants (40–90+ participants each).
Selected Awards
- National Champion, IECEP Interscholastic Quiz Show 2022; Regional 2nd Runner-Up, IECEP-MSC 2022.
- Gold Undergraduate Thesis Nominee, DLSU Department of Electronics and Computer Engineering (2023).
- Team Lead, BPI DataWave Hackathon 2025 (HARAYA cultural-intelligence fintech platform).
- NVIDIA DLI certificates: Fundamentals of Deep Learning (2022); Building LLM Applications with Prompt Engineering (2025).
Skills
- Computer vision and ML: PyTorch, OpenCV, YOLO, NVIDIA DeepStream/TAO, Jetson, CVAT, FiftyOne, scikit-learn
- LLMs and agents: evaluation design, RAG, guardrails, model routing, OpenAI and Gemini APIs, Claude Code, ElevenLabs
- Data and engineering: Python, R, MATLAB, SQL/PostgreSQL, Kafka, AWS, Docker, GitHub Actions, pytest
- Languages: Filipino (native), English
References
Available upon request.