Solutions Engineer
Jun 2023 — PresentSplunk, a Cisco company · Singapore
- A concrete result or responsibility, ideally with a number.
- Another line that shows impact, not just tasks.
About
Solutions Engineer at Splunk (a Cisco company), where I help organizations turn their data into clear, confident decisions — bridging deep technical work with the plain-language storytelling that gets teams to act.
I gravitate toward the pioneer's seat: joining things early, shaping new programs and initiatives from the ground up, and advising others as they find their footing. I care about building things that outlast me and helping the people around me level up along the way.
Experience
Skills
Publications & Projects
Publications
WorldQA: Multimodal World Knowledge in Videos through Long-Chain Reasoning
arXiv preprint arXiv:2405.03272 · 2024
Evaluating Vision-Language Models' Long-Chain Reasoning Ability with Multiple Ground Truths
Final Year Project (thesis), Nanyang Technological University · 2024
Projects
chrisarif.com + AI Concierge
This site — static front end with a Claude-powered concierge, instrumented with OpenTelemetry.
View details →Gatekeeper
A prompt-injection game — outsmart the LLM guarding the gate and talk your way through.
View details →WorldQA Benchmark
Research project benchmarking multimodal world knowledge and long-chain reasoning in video models.
View details →← Back to Publications & Projects
chrisarif.com + AI Concierge
This website — a hand-rolled static site (no frameworks, no build step) hosted on GitHub Pages behind a Cloudflare-managed domain, with a floating AI concierge that answers questions about me.
- Front end: vanilla HTML/CSS/JS with hash-routed views; the chat widget is dependency-free.
- Concierge backend: a small Node/Express proxy that keeps the Claude API key server-side, with CORS allow-listing, per-IP rate limiting, and input sanitization.
- Observability: OpenTelemetry tracing (OTLP/HTTP) with custom spans carrying model and token-usage attributes — pointable at any collector, including Splunk Observability Cloud.
← Back to Publications & Projects
Gatekeeper
A prompt-injection game: an LLM stands guard, and your job is to talk, trick, or inject your way past it. Players probe the model's defenses with social engineering and adversarial prompts — a hands-on (and fun) way to build intuition for LLM security.
Full write-up, tech details, and a playable link coming soon.
← Back to Publications & Projects
WorldQA Benchmark
Research project with S-Lab / MMLab@NTU benchmarking how well multimodal models use world knowledge and long-chain reasoning to answer questions about videos — published as "WorldQA: Multimodal World Knowledge in Videos through Long-Chain Reasoning" (arXiv:2405.03272, 2024).
- The benchmark pairs 1,007 question–answer pairs with 303 videos, requiring models to combine auditory and visual signals with world knowledge to answer correctly.
- Co-authored with Yuanhan Zhang, Kaichen Zhang, Bo Li, Fanyi Pu, Jingkang Yang, and Ziwei Liu.
Achievements
- Award / Milestone — one line of context, and the year.
- Another highlight — what it was and why it mattered.