CLOUD & AI ENGINEER · GOOGLE SINGAPORE

Elias Lim

Cloud and AI engineer experienced in secure cloud architecture, Go and Python backend systems, applied machine learning, and autonomous agent tooling.

Browse projects About me GitHub (@elim316) ↗ LinkedIn ↗ Google Scholar ↗

FEATURED REPOSITORY PREVIEW · LIVE ARCHITECTURE

elim316/Jetski-Agent-Tracer-Plugin Live Execution Graph

CORE ENGINEERING METRICS Verified

Jumpgate Zero-Trust AI Landing Zone < 3 min · $1.96M ARR

Battery SOH Conformal Coverage (IEEE) 95.2% PICP

Meeting Dossier Evaluation Harness 36 / 36 Checks

Scroll down to learn more and explore repositories

ABOUT ELIAS LIM · SINGAPORE

Cloud infrastructure. Backend systems. Applied AI.

BSc (Hons) Computer Science graduate from the University of Glasgow and SIT, building secure cloud architectures, high-scale backend services, and uncertainty-aware AI.

INTERACTIVE ENGINEERING PROFILE · GOOGLE CLOUD SINGAPORE

JUMPGATE DUAL-VPC LANDING ZONE & 14-STEP AGENT VENDING MACHINE < 3 min · 32/32 IM8 · $1.96M ARR

At Google in Singapore, I co-architected Jumpgate, an IM8-compliant Dual-VPC AI agent landing zone and 14-step Vending Machine Agent that compresses 4 to 6 weeks of setup into under 3 minutes (passing 14/14 Ingress and 18/18 Egress security checks), driving $1.96M in realised public sector ARR (+$1.46M pipeline) and training over 4,000 engineers and students across NTU, GovTech, DBS, and A*STAR.

CLICK ANY CARD TO INSPECT CHAPTER Chapter 01 / 04

Jumpgate Dual-VPC & IM8 Security Posture < 3 min · 32 / 32 Checks

Public Sector ARR & Buildathon Reach $1.96M ARR · 4,000+ Trained

KEY HIGHLIGHTS

Get the highlights.

              ZERO-TRUST AGENTIC CLOUD · TERRAFORM & VERTEX AI
              
                Provision IM8-compliant dual-VPC agentic landing zones in under 3 minutes. $1.96M realised public sector ARR.
              
                INGRESS VPC · DMZ
                Cloud Armor WAF + L7 LB
                14 / 14 Runtime Checks
                Zero Public IPs

                EGRESS VPC · BRAIN
                Private Service Connect
                18 / 18 Static Checks
                Vertex AI & Model Armor

                14-STEP ADLC
                >=0.85 Judge Gate
                < 3 min Setup
                $1.96M ARR
              
              32 / 32 Security Checks · 70% Less Engineering Effort
              Explore Jumpgate Stage ↓

              AGENT OBSERVABILITY · GO, SVELTE 5 & CONNECT-RPC
              
                Inspect multi-agent execution graphs, tool latency, and 128k context window saturation in real time.
              
                JSONL WATCHER
                fsnotify Tail
                Zero Cloud Lock-In

                CONNECT-RPC STREAM
                Planner · Tool · Subagent
                Sub-ms Span Diffs

                128K CONTEXT
                Token Burn Telemetry
                Live Cost Estimator
              
              Native Sidecar · Real-Time DAG & Span Inspector
              Explore Tracer Stage ↓

              FIRST-AUTHOR IEEE XPLORE · CONFORMAL PREDICTION & XAI
              
                95.2% empirical coverage (0.012 ECE) for battery health and 64.96% smaller prediction sets in medical imaging.
              
                ADAPTIVE CONFORMAL
                95.2% Empirical PICP
                0.012 Calibration ECE

                SHAP · LIME · GRAD-CAM
                Feature Attribution
                Voltage & Temp Drivers

                MEDICAL XAI
                90.90% Coverage
                -64.96% Set Size
              
              APSIPA ASC 2025 · IEEE Xplore Doc 11249263
              Explore Research Stage ↓

              TECHNICAL ENABLEMENT & NATIONAL HACKATHON HONOURS
              
                4,000+ engineers and students trained across GovTech, DBS, NTU, and A*STAR, plus 3x National Award Winner.
              
                4,000+ ENGINEERS & STUDENTS
                GovTech · DBS Bank · NTU · A*STAR
                Agentic AI & Cloud Buildathons

                3X NATIONAL HACKATHON HONOURS
                2nd Place NUS/Singtel LifeHack 2025
                Overall Best AISG/SMU · DSTA TIL-AI
              
              Enterprise Workshops · Competitive Systems Engineering
              Explore Honours & Enablement ↓

ZERO-TRUST AGENTIC CLOUD ARCHITECTURE · TERRAFORM & VERTEX AI

Jumpgate · Agentic AI Landing Zone & Vending Machine

IM8-compliant Dual-VPC landing zone and 14-step Vending Machine Agent that compresses 4 to 6 weeks of cloud setup into under 3 minutes, driving $1.96M in realised public sector ARR (+$1.46M pipeline).

View Landing Zone Repo ↗ View Vending Machine Repo ↗

JUMPGATE DUAL-VPC & 14-STEP VENDING MACHINE · INTERACTIVE TOPOLOGY

ZERO-TRUST TOPOLOGY (CLICK ANY NODE) 14-Step ADLC · < 3 min (>=0.85 Judge)

IM8 POSTURE & COMMERCIAL IMPACT 32 / 32 Checks Passed

Ingress Runtime Posture (WAF, TLS, Serverless NEG) 14 / 14 PASS

Egress Static Terraform Checks (PSC Vertex AI, SWP L7) 18 / 18 PASS

Public Sector Realised & Pipeline ARR $1.96M + $1.46M

AGENTIC OBSERVABILITY & WORKSPACE · PYTHON & VIS.JS

Jetski Agent Tracer & Harness

Visualise every step an AI agent takes in real time, from user prompt and LLM planning to tool execution and subagent delegation, paired with a 2x2 workspace that monitors 200k token context usage.

View Agent Tracer Repo ↗ View Jetski Harness Repo ↗

JETSKI AGENT TRACER & HARNESS · INTERACTIVE ARCHITECTURE

AGENT EXECUTION GRAPH (CLICK ANY STEP) Step #02 PLANNER · 640ms

2x2 WORKSPACE & CONTEXT WINDOW METER 142k / 200k · 71% Cached

MULTI-CORPUS AGENT & EVALUATION HARNESS · PYTHON

Smart Meeting Prep & Dossier Agent

Two-stage scheduled agent that synthesises Calendar, Gmail, Chat, Drive, and People Directory context into cited one-page briefings, verified by a 5-scenario, 36-check hallucination linter.

View Repository ↗ Technical specs ↓

MEETING PREP & DOSSIER AGENT · EVALUATION HARNESS

TWO-STAGE SCHEDULED PIPELINE Stage 2 · Hallucination Linter (36/36)

5-SCENARIO MOCK EVALUATION SUITE Zero Hallucinated URLs

Citation & Source URL Verification 100% Grounded

Cross-Corpus Synthesis (Calendar/Gmail/Chat/Drive) 5 / 5 Scenarios

Stateless 60-Min Window Partitioning Zero Duplicates

FIRST AUTHOR · APSIPA ASC 2025 & IEEE XPLORE (DOC 11249263)

Uncertainty & Explainable AI for Battery Analytics

Unified deep learning framework quantifying both model and data uncertainty via Adaptive Conformal Inference (ACI), calibration metrics (PICP, ECE), and SHAP/LIME attributions.

View Repository ↗ Read IEEE Paper ↗

A*STAR BATTERY SOH FRAMEWORK · CONFORMAL & XAI SANDBOX

CONFORMAL BAND (95% ACI) PICP: 0.952 · ECE: 0.012

CALIBRATION & POST-HOC EXPLAINABILITY SHAP · LIME · IG

Prediction Interval Coverage (PICP @ 95%) 95.2% Empirical

Expected Calibration Error (ECE Reduction) 0.012 ECE

Feature Attribution Consistency (SHAP / LIME) Voltage & Temp

ALL REPOSITORIES

Explore the full project line-up.

Filter by engineering domain or expand any project tile to inspect its technical architecture.

SIDE-BY-SIDE ARCHITECTURE COMPARISON

Compare systems.

Select any two repositories to compare their runtime stack, verification gates, and system schematics.

System A System B

EXPERIENCE & EDUCATION

Career and academic progression.

Engineering roles across cloud infrastructure, high-scale backend systems, and applied machine learning research.

Apr 2026 to Present Google · Singapore

Customer Engineer · Google

Co-architected Jumpgate, an IM8-compliant Dual-VPC AI agent landing zone (Terraform, Cloud Armor WAF, Serverless NEG, Private Service Connect, Secure Web Proxy) and 14-step Vending Machine Agent (with >=0.85 LLM-as-a-Judge quality gate) that cuts deployment from 4 to 6 weeks to under 3 minutes while passing 14/14 Ingress runtime and 18/18 Egress static security checks. Contributed to $1.96M in realised public sector ARR (plus $1.46M pipeline ARR), built the native Jetski Agent Tracer visualiser, and led AI buildathons for 4,000+ engineers and students.

Jan 2026 to Mar 2026 Grab · Singapore

Trust Software Engineer · Grab

Built scalable Golang backend services for real-time fraud detection, implementing runtime control mechanisms such as feature flags and rate limiting on petabyte-scale risk infrastructure. Expanded automated unit test coverage, resolved production defects, and contributed to architecture and code reviews across Agile sprints.

Sep 2024 to May 2025 A*STAR · Singapore

AI Research Intern · A*STAR (Agency for Science, Technology and Research)

Designed and implemented a unified uncertainty-aware deep learning framework for battery State-of-Health estimation that quantifies both model and data uncertainty alongside model-agnostic explainability (SHAP and LIME). First author of "A Unified Framework for Interpretable and Uncertainty-Aware Battery State of Health Estimation Using Deep Neural Networks" (APSIPA ASC 2025 / IEEE Xplore).

Aug 2022 to May 2025 Glasgow & SIT

BSc (Hons) Computer Science (Second Upper Class) · University of Glasgow & SIT

Awarded 2nd Place Overall at NUS, Singtel, and Millennium Management LifeHack 2025 (EduVerse), Overall Best (Intermediate) and Most Impactful at the AI Singapore, SMU, and CyberYouthSG Hackathon, and Semifinalist at DSTA BrainHack 2024 TIL-AI (Autonomous Air Defence System).

NATIONAL HACKATHON HONOURS 3x Award Winner / Finalist

Competitive AI & Systems Engineering

TECHNICAL ENABLEMENT & WORKSHOPS 4,000+ Engineers & Students

AI Buildathons & Cloud Architecture Training

Designed and delivered hands-on agentic AI, Vertex AI, and zero-trust Google Cloud architecture buildathons across Singapore's public sector, enterprise banking, and research institutions.

GovTech Singapore DBS Bank NTU Singapore A*STAR Singapore

PUBLICATIONS

Peer-reviewed research.

Published in IEEE Xplore, with one-click BibTeX citations.

APSIPA ASC 2025 · IEEE Xplore (Doc 11249263) First Author · 2025

A Unified Framework for Interpretable and Uncertainty-Aware Battery State of Health Estimation Using Deep Neural Networks

Designed a unified deep learning framework quantifying both model and data uncertainty via Adaptive Conformal Inference (ACI) and calibration metrics (PICP, ECE, MCE), paired with model-agnostic explainability (SHAP, LIME, Integrated Gradients) across the McMaster and Oxford battery datasets.

IEEE Paper Code

IEEE Xplore · Document 10892827 2025

Teamwork Assessment in Software Engineering Education

Investigated data-driven approaches for evaluating collaboration and student team dynamics in software engineering coursework.

IEEE Paper

SPECIFICATIONS & TELEMETRY

By the numbers. Built for production.

Compare the production architecture against legacy baselines, or click any tile below to filter repositories.