AI Engineer

SENIOR AI ENGINEER

COMPANY: EKLIPSE

ABOUT THE ROLE

As a Senior AI Engineer at Eklipse, you will play a pivotal role in advancing our AI-powered highlight detection and content intelligence systems that serve gaming creators worldwide.

This is a full-stack AI position spanning the entire lifecycle — from research and experimentation with state-of-the-art models, to building production-grade pipelines at scale.

You will work at the intersection of multimodal AI (video, audio, text), modern LLMs and

VLMs, agentic systems, and production ML infrastructure. You will collaborate closely with the CTO, product, and engineering teams to translate business problems into scalable AI solutions. Beyond shipping models, we expect you to actively leverage AI tools in your own workflow to multiply your output and help the team move faster.

KEY RESPONSIBILITIES

• Build and maintain multimodal pipelines that jointly reason over video, audio, gameplay context, and text for integrated content analysis.
• Design and develop agentic AI systems using modern frameworks (LangGraph, CrewAI, AutoGen, or custom orchestration), implementing the Model Context Protocol (MCP) for standardized tool use and external integrations.
• Integrate AI solutions into production using microservices, API gateways, and event-driven processing, supported by model serving infrastructure (vLLM, SGLang,TensorRT-LLM, TGI, vector databases.
• Optimize inference through quantization (GPTQ, AWQ, GGUF), knowledge distillation, and speculative decoding; build evaluation frameworks, observability stacks, and guardrails covering prompt injection detection, output filtering, PII protection, and content safety.
• Serve as a technical mentor for junior engineers and communicate complex technical concepts clearly to non-technical stakeholders, including executive leadership.
• Own and evolve the multimodal highlight detection pipeline end-to-end — video, audio, transcript, and gameplay context fusion across engine tiers. Make signal routing and compute allocation decisions.
• Drive VLM evaluation-to-production lifecycle: prompt engineering, failure-mode analysis, go/no-go gating. Own fine-tuning pipelines on proprietary data (dataset curation, SFT/RL, experiment tracking).• Design agentic AI systems using modern frameworks (LangGraph, CrewAI, AutoGen, or custom orchestration), implementing MCP for standardized tool use and external integrations.
• Build game-specific evaluation harnesses with per-game, per-error-type failure decomposition. Maintain benchmark datasets and enforce phase-gate criteria before changes reach production.
• Integrate AI components into production using microservices, API gateways, and event-driven processing. Build observability, guardrails (prompt injection, output filtering, content safety), and graceful degradation paths. Close the data flywheel: instrument production predictions, collect correction signals, grow the proprietary labeled dataset, and feed it back into fine-tuning and eval.
• Work fluently with video processing — shot detection, codec-aware chunking, frame extraction, quality profiling — and make tradeoffs between visual fidelity, chunk size, and model performance.
• Mentor junior engineers and communicate pipeline performance, accuracy tradeoffs, and infrastructure costs clearly to non-technical stakeholders and executive leadership.

REQUIREMENTS

• Bachelor's or Master's degree in Computer Science, AI, Machine Learning, or a related field (or equivalent practical experience).
• English for Technical sharing
• Minimum 5 years developing and deploying AI solutions in production, with at least 2 years focused on LLMs, VLMs, or generative AI.
• Demonstrable portfolio of production-grade systems.
• Hands-on experience fine-tuning and deploying modern LLMs/VLMs. Comfortable with SFT/RL workflows, dataset curation, and experiment tracking.
• Proficiency in agentic AI development: state management, tool calling, multi-agent orchestration, and memory systems. Advanced prompt engineering skills (fewshot structured outputs, constrained decoding).
• Experience with multimodal AI: video understanding, audio signal processing, cross-modal fusion, and related preprocessing techniques. Familiarity with video pipelines, frame extraction, or media processing at scale.
• Experience designing custom evaluation frameworks beyond standard benchmarks - failure-mode decomposition, domain-specific test sets, go/no-go gating, and regression tracking.
• Track record of optimizing inference costs and managing compute budgets -quantization, batching strategies, spot instance orchestration, or adaptive compute allocation.
• Strong technical writing and communication skills. Able to translate model performance, accuracy tradeoffs, and infrastructure costs into clear narratives for on-technical stakeholders.
• Superior problem-solving and analytical thinking. Comfortable operating with ambiguity and making tradeoff decisions with incomplete data.

PREFERRED QUALIFICATIONS

• Experience with reasoning models (o3/o4-class, DeepSeek-R1, QwQ) and testtime compute scaling techniques.
• Background in gaming, streaming, or creator-economy products.
• Experience building real-time video analysis pipelines (FFmpeg, GStreamer) or on-device inference (ONNX, CoreML, mobile deployment).

INTERVIEW PROCESS: 2-3 rounds

WORKING TIME: Mon-Fri (9am-6pm). This is Fulltime Position

WORKING PLACE: The Miss Ao Dai Building, Floor 10th, 21 Nguyễn Trung Ngạn, Bến Nghé, Quận 1, HCMC