Trajectory Raises $45M at $300M Valuation to Let Non-Technical Experts Train AI Models

The barrier to AI model training is being pushed even lower by capital and startups. Trajectory, a startup founded just one year ago, has announced $45 million in funding at a company valuation of $300 million. The capital will be used to enable domain experts without technical backgrounds to participate in the AI model training process — potentially a key step as AI applications move into deeper waters.

Funding Overview: $300M Valuation in One Year

Founded just a year ago, Trajectory has secured $45 million in funding at a $300 million valuation. This pace is not unusual in the current AI startup environment, but it also reflects strong capital market interest in the direction of "democratizing AI training."

Judging by the ratio of funding to valuation, investors are clearly not looking at short-term revenue but betting on a longer-term trend: the primary actors in AI model training are expanding from a small number of algorithm engineers to a large pool of experts who possess domain knowledge but cannot write code.

Why "Non-Technical Experts Training AI" Matters

Traditionally, training and fine-tuning AI models has been the exclusive domain of machine learning engineers. Data labeling, model fine-tuning, and evaluation iteration require understanding concepts such as loss functions, learning rates, and overfitting. But in industries like healthcare, law, finance, and manufacturing, the people who truly understand the business logic often do not write code.

Trajectory targets precisely this gap. If the AI industry over the past few years has solved the "model capability" problem, what comes next is solving the "adaptation of models to specific business scenarios" problem. And the latter precisely requires deep involvement from non-technical experts.

According to the original report, Trajectory's positioning is to enable non-technical experts to participate in AI model training. This means it may provide some kind of abstraction layer that encapsulates complex training processes into interfaces or workflows that domain experts can operate.

Loss functions, learning rates, and overfitting are three fundamental concepts in machine learning training. A loss function measures the deviation between model predictions and actual results, and the goal of training is to minimize this deviation; the learning rate controls the step size of each parameter update — too large and it tends to oscillate and fail to converge, too small and training is slow; overfitting refers to a model performing well on training data but having poor generalization on new data. The key value of platforms like Trajectory is to encapsulate the process of tuning these hyperparameters, so that domain experts only need to focus on data quality and business objectives without understanding the underlying mathematical mechanisms.

Where the Money Will Go: Productization and Market Education

$45 million is a considerable scale for early-stage funding. For a company founded just one year ago, this capital will most likely go toward three directions: product R&D, team expansion, and market education.

Market education may be underestimated. Getting non-technical experts to accept the idea that "I can also train AI" takes time. This is not simply a product experience issue but a reshaping of workflows and organizational habits. Trajectory needs to prove that models trained with domain expert involvement genuinely outperform versions built in isolation by purely technical teams in terms of business performance.

AutoML (Automated Machine Learning) refers to technologies that automate processes such as feature engineering, model selection, and hyperparameter tuning, with the goal of lowering the barrier to machine learning applications. It overlaps with Trajectory's positioning but differs in focus: AutoML mainly addresses automation of "selecting models and tuning parameters," while Trajectory emphasizes enabling domain experts to participate in constructing training data and shaping model behavior. Understanding this difference helps assess the competitive boundaries of this track — AutoML tools often still target users with some technical foundation, while Trajectory attempts to move the entry point further upstream to purely business roles.

Competitive Landscape and Potential Challenges

" enabling non-technical people to use AI" is not an entirely new track. Low-code/no-code AI platforms, AutoML tools, and various large model fine-tuning services are all approaching the same goal from different angles. Trajectory's differentiation lies in its emphasis on "training" rather than just "using" — a more upstream position with higher barriers.

The challenges are equally clear. First, data security and compliance, since domain experts often handle core enterprise data; second, evaluation criteria for training effectiveness — how can non-technical experts judge whether a model is "trained well"; third, how to avoid oversimplifying complex problems, which could lead to unreliable models in real-world scenarios.

Large model fine-tuning refers to continuing to train a pre-trained large model with domain-specific data to make the model better suited to a particular task. Compared with training from scratch, fine-tuning costs far less, making it the mainstream approach for enterprises to implement AI. The effectiveness of fine-tuning depends heavily on data quality and labeling consistency — precisely where domain experts can leverage their strengths. If Trajectory can productize the fine-tuning process, letting experts describe requirements in business language while the platform handles technical implementation, it can occupy a unique ecological niche between "using AI" and "training AI."

What It Means for the AI Industry

Trajectory's funding is a signal: competition in AI infrastructure is shifting from computing power and the model layer to "who can enable more people to participate in model building." If this path succeeds, the supply of AI models will no longer be limited by the number of engineers but by the number of domain experts — a far larger pool.

Of course, a $300 million valuation after one year also implies high expectations. Trajectory will need to prove going forward that models trained by non-technical experts can truly be deployed in practice, not just remain at the demo stage.

One-line Summary

Trajectory, founded one year ago, raises $45M at a $300M valuation, betting on non-technical experts training AI.

Paragraph Summary

Trajectory, a startup founded just one year ago, has announced $45 million in funding at a $300 million valuation. The company's core positioning is to enable non-technical experts in fields such as healthcare, law, and finance to participate in AI model training, bridging the gap between business knowledge and algorithm engineering. The article argues that the funds will mainly go toward product R&D, team expansion, and market education, with market education potentially being the underestimated challenge. The track faces challenges including data security and compliance, the lack of evaluation standards for training effectiveness, and unreliable models caused by oversimplification. Trajectory's differentiation lies in emphasizing "training" rather than merely "using," occupying a more upstream position with higher barriers. This funding is seen as a signal that competition in AI infrastructure is shifting from computing power and the model layer to "the scale of those who participate in building."