Founding AI Engineer

About the Role

TaloTrace is pioneering AI agents that autonomously test software on real devices by exploring web apps and mobile builds, identifying expected behavior, uncovering bugs, and reporting them across platforms including Web, Android, and iOS on both emulators and physical hardware. As we operate with a small, driven team moving at a rapid iteration rate, we seek a Founding AI Engineer in Singapore to take full ownership of the model layer powering these agents. The models enable the agents to interpret screens, decide actions, perform them, and assess outcomes, with each step potentially refined for speed, cost, size, and accuracy improvements.

Core Responsibilities

• Manage the entire model pipeline—from data collection, training, evaluation, to deployment and operational costs.
• Translate research findings into robust production systems, discontinuing models that do not meet standards.
• Develop truthful evaluation frameworks, including recognizing when results are inconclusive or fail.
• Balance tradeoffs among cost, performance quality, and latency with data-driven decisions.
• Establish company-wide standards for model development including which metrics and improvements warrant deployment.

Potential Project Focus Areas (Pick 2-3 Initially)

• Creating smaller models replacing larger ones using techniques like distillation, fine-tuning, quantization for efficient computation.
• Developing reliable grounding and perception models for element localization across diverse platforms and after UI changes.
• Enhancing judgment layers to assess the validity, severity, and clarity of findings.
• Designing exploration policies that mimic natural user navigation paths rather than just ideal scenarios.
• Predicting screen state changes post actions to improve navigation and anomaly detection.
• Building evaluation infrastructure featuring offline benchmarks, regression gates, and trustworthy metrics.
• Optimizing inference costs with quantization, batching, on-device deployment, and cache mechanisms.
• Establishing and maintaining training data pipelines including collection, cleaning, labeling, and version control.

Candidate Qualifications

• Over 4 years of experience in machine learning or AI engineering, or strong software engineering skills with significant practical model ownership in production.
• Advanced proficiency in Python and practical experience with PyTorch or similar frameworks, capable of reproducing research codebases.
• Hands-on experience training or improving models through supervised fine-tuning, LoRA/PEFT, reinforcement learning, behavior cloning, or distillation.
• Skilled at constructing evaluation mechanisms that reliably identify valid metrics and detect faults.
• Demonstrated production experience working with LLM systems, including structured output generation, tool integration, and managing latency and costs.
• Comfortable working with data pipelines for gathering, cleansing, and versioning training datasets.
• Methodical in verifying results before reporting and transparent about failures.
• Capable of rapid prototyping to explore hypotheses and willing to discard non-viable solutions.
• Excellent written communication to document experiments for team awareness and continuation.
• Mastery of AI coding tools with consistently high-quality deliverables.

Desirable Additional Experience

• Early engineering experience in startups, familiar with launching initial product versions.
• Experience with vision-language or multimodal model training.
• Work related to GUI agents, computer task automation, robotic policies, or embodied AI.
• Expertise in imitation learning or behavior cloning from human demonstration data.
• Optimization of inference via advanced quantization methods (FP8, INT4), vLLM, TensorRT, KV-cache, or on-device deployment.
• Experience with reinforcement learning techniques such as RLHF, GRPO, or reward-verifiable systems.
• Modeling human behavior in domains like game AI, recommender systems, user simulation, or behavioral biometrics.
• Contributions to research through publications, open source projects, or replication of relevant work.

Application Process

The selection process includes a 7-day coding challenge followed by an interview with the CTO.

Why Join TaloTrace?

Our commitment is to deliver quality software solutions that empower the next generation of creators and innovators to reach broader audiences effortlessly. We rapidly validate model improvements using real customer runs, supporting swift iteration cycles measured in days rather than quarters. We foster an inclusive atmosphere welcoming all applicants and accommodate individual needs through the hiring process.

Company Culture

Our team is small and moves fast with minimal formalities. Decisions are driven by experiments rather than debates when possible, with open discussions on direction, priorities, and quality standards. We value constructive disagreements and encourage team members to extend beyond their primary roles.