Trajectory reaches $45M: the new AI unicorn for non‑experts | Antonio Pedoto
Among the startup Trajectory, the new $45 million round from Sequoia, Accel, and Google Cloud Ventures pushed the valuation to $300 million, promising an AI pla
The Trajectory Leap: From Startup to Unicorn in One Year
The latest financing round of $45 million made it clear that Trajectory is no longer just an experiment, but a reality ready to overturn the rules of the game. The capital was led by Sequoia Capital and Accel, two venture‑capital giants already known for investing in Snowflake and Databricks. Completing the round is also Google Cloud Ventures, which highlighted interest in the platform’s ability to “break down barriers between data science and business.” The composition of investors is not accidental: funds with AI‑infrastructure experience and partners with global sales networks indicate an aggressive scaling strategy and the desire to position Trajectory as “the entry point” for organizations that want to experiment with AI without a dedicated engineering team.
This liquidity injection pushed the valuation to $300 million, placing the startup among the youngest unicorns in the AI sector. To put it in perspective, compare Trajectory with competitors like DataRobot (valuation around $2.5 billion) and H2O.ai (about $1.2 billion). While these giants boast a larger customer base and a longer history, Trajectory has gained ground thanks to two key levers: a drag‑and‑drop UI that lets a product manager assemble a classification model in under 15 minutes, and a serverless architecture that eliminates the need to manage GPUs or Kubernetes clusters. In terms of valuation‑to‑years‑in‑business ratio, Trajectory ranks among the most efficient startups on the scene: a $300 M value in just 12 months of existence is rare, especially in a market where the average is 3–5 years to reach the $100 M threshold.
The growth strategy is equally bold. Trajectory aims to triple its customer base within the next 18 months, leveraging three operational levers: Partner channel – thanks to the involvement of Google Cloud Ventures, the startup is integrating its marketplace with Google Cloud Marketplace, allowing companies to spin up a Trajectory instance with a click and pay “pay‑as‑you‑go.” Rapid verticalization – the product team has already released pre‑trained models for sectors such as retail (product recommendation) and healthcare (symptom triage). These “turn‑key” solutions reduce time‑to‑value for non‑technical customers.
Certification program – a series of certified online courses, sponsored by educational partners, aim to create a community of “AI champions” within traditional enterprises.
💡 Did you know... Trajectory has already signed pilot contracts with more than 30 multinationals, including a well‑known fashion brand that used the platform to optimize seasonal demand forecasting, cutting excess inventory by 12 % in just three months.
The result is a market positioning that combines accessibility (no‑code) and scalability (cloud‑native architecture). In an ecosystem where technical complexity remains the main barrier to AI adoption, Trajectory has shown that it’s possible to scale quickly, attract high‑level capital, and, most importantly, turn an “AI for everyone” idea into a $300 M business reality in less than a year.
Why AI Training Still Remains a Privilege for Tech Experts
Artificial intelligence has promised to revolutionize every industry, yet today its implementation remains a privilege reserved for the few. Despite advancements in low-code solutions and open-source frameworks, training a high-quality AI model still demands advanced skills in programming, data engineering, and managing complex infrastructures. Consider a marketing team aiming to fine-tune a natural language processing model to analyze customer feedback in real time: to achieve this, they’d first need to build a pipeline for clean data, define a GPU-based training architecture, optimize hyperparameters, and handle deployment. Without a team of data scientists or ML engineers, the project quickly turns into a nightmare of dependencies and hidden costs.
The complexity isn’t just technical—it’s also organizational. A company deciding to develop an in-house model faces a series of structural hurdles: Specialized skills: Writing an efficient data loader in Python or configuring a Kubernetes cluster for training takes months of training. Even tools like Hugging Face or TensorFlow don’t eliminate the need to grasp concepts like tokenization, fine-tuning, or batch normalization. Expensive infrastructure: Training a large-scale model on AWS or GCP can cost thousands of dollars per day just for GPU usage. Without a dedicated budget or cloud access, many companies are forced to abandon the idea or settle for pre-trained models—often ill-suited to their specific use cases. Data engineering: 80% of the work in an AI project goes into data preparation. Cleaning datasets, managing accurate labeling, and ensuring reproducibility require tools like Pandas, PySpark, or advanced SQL, along with deep expertise in data pipelines.
🎯 Practical tip: If your team lacks internal expertise, consider using pre-configured notebooks on Google Colab or Kaggle, which offer free GPUs for initial experiments. However, remember these tools are limited to small or medium-sized models: for production projects, complexity inevitably resurfaces.
The gap between those who can train AI models and those who can’t isn’t just technical—it’s cultural. Marketing, operations, and customer support teams often find themselves in a subordinate position to IT departments, which hold a monopoly over resources and expertise. This creates a vicious cycle: non-technical departments must wait for engineers to implement their ideas, losing agility and competitiveness. The example of a company forced to wait six months for a custom sentiment analysis model—despite having ready-made datasets—highlights the issue: AI becomes a bottleneck instead of an accelerator.
The solution isn’t just democratizing access to tools but redesigning processes so that AI model creation is integrated into daily workflows—not a standalone project reserved for specialists.
How Trajectory’s No-Code Platform Works
Trajectory designed its suite as a modular block-based system. Every stage of an AI model’s lifecycle—from data ingestion to deployment—is accessible via a drag-and-drop interface. No more lines of complex code, no advanced programming skills required.
Let’s be clear: the goal is to turn tasks that typically take months of work by data scientists and ML engineers into just a few clicks and guided configurations.
At the core of the architecture lies a visual pipeline divided into four key phases: Data Ingestion, Preprocessing, Training, and Deployment. Each module connects seamlessly to the next; data flows automatically from one stage to another without requiring custom scripts.
Imagine, for example, a marketing team uploading a CSV of customer reviews, applying cleaning filters (removing null values, tokenization) through a simple dropdown menu, and launching an emotion-classification model with a single click. The result? An entire workflow ready in minutes.
AutoML and Automatic Optimization Trajectory integrates AutoML (Automated Machine Learning) techniques. The system instantly analyzes dataset characteristics—size, distribution, variable types—and suggests the most suitable models, whether neural networks, decision trees, or transformer-based architectures.
Here’s the kicker: the hyperparameter tuning algorithm finds optimal configurations, cutting manual work by 70–80% compared to traditional methods.
💡 Did you know... Trajectory also supports fine-tuning on pre-trained models (like those from Hugging Face), letting you customize large language or visual models without worrying about underlying infrastructure.
Scalability and Security: Cloud and On-Premise The platform runs seamlessly on public cloud infrastructures (AWS, GCP, Azure) or on-premise. This means flexibility for compliance needs or keeping data under strict control.
Users can choose from various scalability options—dedicated GPUs, distributed clusters—and none of it requires manual environment setup.
Step back: a healthcare company could use Trajectory on-premise to train an AI-assisted diagnosis model, keeping data within its network and complying with GDPR. Meanwhile, a tech startup could leverage AWS SageMaker integration to scale training across hundreds of GPU instances during peak demand.
A Practical Example: From Data Collection to Deployment Ever wondered how an e-commerce team could predict customer churn? With Trajectory, the process is straightforward: Data Ingestion – Load a historical dataset of transactions and user behavior, whether from an SQL database or a CSV file. Preprocessing – Normalize features, handle missing values, and create new variables (e.g., “days since last purchase”) via a menu-driven interface. Training – The platform automatically selects a model (XGBoost, neural network, etc.) and optimizes hyperparameters with a guided wizard. Deployment – The model is published as a REST API or embedded in a CRM, with continuous performance monitoring.
All of this—without writing a single line of Python or R. The result? A production-ready, scalable, and reliable model.
So, what’s the hold-up? Isn’t it time to try a solution this streamlined?
Real‑world use cases: from customer support to demand forecasting
The true value of Trajectory reveals itself in its practical use cases, where non‑technical teams turn their data into high‑performing AI without writing a single script. Here’s how the platform is already changing the way things are done in three key sectors. Custom chatbots for customer support
Imagine an e‑commerce company that wants to cut response times and boost customer satisfaction. With Trajectory, the Customer Experience team imports chat histories, emails, and tickets in CSV format, defines basic intents (e.g., “Refund request”, “Shipping issue”) and launches an NLP model in a few clicks. The platform automatically generates the tokenization, training, and tuning pipeline, delivering a chatbot ready to be integrated with the ticketing system. In practice, the department no longer depends on data‑science engineers; it can test different response configurations, add new FAQs, and monitor accuracy metrics through Trajectory’s intuitive dashboard. Churn‑prediction models for marketing teams
A SaaS company with a customer base of 12 000 users wants to know who is most likely to cancel their subscription. The marketing lead imports usage logs, demographic data, and support interactions. Trajectory creates a binary classification model, automatically optimizing feature‑engineering and model parameters. The result is an output file that shows churn risk for each user, with a graphical interface that lets you filter by market segment and export the results in a format compatible with the CRM. The team can then launch targeted retention campaigns without writing SQL or Python. Supply‑chain optimization through AI‑based forecasting
A manufacturer of electronic components needs to forecast demand for the next 12 months to avoid both overproduction and stockouts. The logistics department imports historical sales data, seasonality, and promotional events in a CSV file. Trajectory uses a time‑series model, such as Prophet or LSTM, and trains it automatically, selecting the best configuration based on forecasting quality. The result is a JSON‑formatted forecast that can be integrated directly into the ERP system via a simple REST endpoint. With this visibility, the supply‑chain team can plan raw‑material orders with a safety margin reduced by 30 %, improving efficiency and cutting costs.
💡 Did you know: according to industry studies, companies that adopt low‑code AI platforms for churn prediction see an average 15 % increase in retention rate within six months of implementation.
🎯 Practical tip: start with a pilot project in an easy‑to‑understand domain (e.g., customer support) to demonstrate value to decision‑makers, then scale to more complex use cases like supply chain.
In short, Trajectory shows that accessibility to machine learning is no longer a luxury reserved for data scientists. With a few clicks, professionals can build, test, and deploy AI models that deliver measurable impact in a short time.
Risks and Limits: When Simplicity Hides Invisible Traps
Trajectory promises to democratize the training of artificial intelligence models, but there’s one detail that cannot be ignored: ease of use is not synonymous with safety. In fact, it can sometimes become a problem. The subtlest risk? Believing that a single click is enough to obtain reliable results, without understanding what’s really behind it. There are three scenarios where a lack of technical skills can turn into a ticking time bomb: hidden biases in the data, model vulnerabilities, and the wide gaps in compliance.
Systemic Bias: When AI Repeats Errors No One Sees
An artificial intelligence model is nothing more than a mirror: it reflects what it is shown. The problem? If the person training it doesn’t know what they’re looking at, they can end up amplifying prejudices without realizing it. All it takes is an unbalanced dataset, a non‑diverse source, or worse, a sample of data that reproduces historical imbalances—and the model automatically perpetuates them.
Trajectory provides tools to customize models, but without at least a basic knowledge of statistics and data science, the average user could end up accepting everything the platform proposes. An example? A recruiting algorithm trained on outdated company data could penalize female candidates or those from minority ethnic backgrounds, simply because historical numbers suggest it. And if the user doesn’t know how to balance the dataset? The bias becomes invisible, but not any less harmful.
To be blunt, the problem isn’t just theoretical. A 2023 MIT study found that 76 % of AI‑based hiring models exhibited discrimination risks, often because the datasets hadn’t been properly cleaned or diversified. A data scientist would have spotted the discrepancies; an inexperienced user might just hit “Run” and hope for the best.
Adversarial Attacks: When a Few Pixels Are Enough to Throw Everything Off
The security of a model trained with no‑code tools is another sore spot. An adversarial attack can be enough: slightly altered inputs—a picture with almost imperceptible pixel changes, a text with barely tweaked characters—can cause the model to make serious errors. The problem? Someone without technical skills doesn’t even know the risk exists.
Here are some measures an inexperienced user might overlook: Input filters to block suspicious data. Cross‑validation to test model robustness. Sandbox isolation (like a Docker container) to limit damage in case of an attack.
This code shows how to isolate a model in a controlled environment. But if the user doesn’t know what they’re doing? The model could be exposed to overload, attacks, or malfunctions.
Compliance: When the Law Doesn’t Stop at the Interface
Then there’s the regulatory side. GDPR, CCPA, sector‑specific regulations: all laws that demand traceability, transparency, and user rights. But if the person training the model lacks technical expertise, how can they guarantee compliance?
Common mistakes include: Failing to document data sources or applied changes. Ignoring auditability, making it impossible to prove compliance during an inspection. Underestimating the right to explanation***, i.e., users’ right to ask for reasons behind automated decisions.
The point is: even the most intuitive platform can’t replace basic knowledge. If every step—from the datasets used to the modifications applied—is not tracked, compliance becomes a façade.
🎯 A tip (that’s not obvious): Document everything as if you had to explain it to a regulator. Use automatic logging tools, even simple ones like these:
A basic setup to log operations. Without something like this, compliance is just an empty word.
The Real Question: Who Takes Responsibility?
Trajectory is changing the rules of the game, but simplicity cannot be an excuse for neglecting responsibility. The risk is not only technical but also ethical and legal. The challenge isn’t just making AI accessible to everyone; it’s ensuring that those who use it know what they’re doing—or at least have the tools to avoid ending up with a model that discriminates, gets hacked, or breaks the law.
In the end, artificial intelligence doesn’t become democratic just because it’s easy to use. It becomes responsible when its users understand—or at least get help to understand—what lies behind the “Run” button.
Implications for IT Professionals and Future AI Market Outlook
The arrival of Trajectory marks a turning point in the distribution of AI skills within organizations. The no‑code platform does not replace data scientists; it reconfigures their role, creating a new hierarchy that demands orchestration, governance, and monitoring skills more than pure programming.
New Skills Required With a drag‑and‑drop interface, product, marketing, or HR teams can define training workflows: select datasets, choose a pre‑trained model, set hyper‑parameters, and launch training with a click. The IT professional becomes a pipeline orchestrator. They must know how to: Configure and version workflows – understand how to save a pipeline definition in an internal Git repository or a process‑management tool (e.g., Apache Airflow) to ensure reproducibility. Implement governance controls – define policies for accessing sensitive data, set GPU‑usage spending limits, and verify compliance with regulations such as GDPR. Monitor performance metrics in real time – use integrated dashboards to track loss, accuracy, and data drift, intervening quickly when models show signs of degradation.
Evolution of the Data Scientist Demand The traditional “data scientist‑developer” profile is experiencing a saturation of repetitive tasks. Basic preprocessing, standard architecture selection, and automatic tuning will be handled by the platform. Consequently: Greater focus on problem framing: data scientists will spend more time defining the business problem, interpreting results, and translating insights into operational actions. Specialization in advanced MLOps: managing production environments, scaling on hybrid cloud, and integrating with legacy systems will become the most valuable areas. Role of “AI coach”: experts who support non‑tech teams, translating business needs into workflow configurations and ensuring model quality.
A practical example: the customer‑care department of a mid‑size e‑commerce company uses Trajectory to build a support‑request classification model. The support team defines the workflow, uploads chat logs, and in minutes obtains a model ready for CRM integration. The company’s data scientist intervenes only to validate metrics and set a drift alert, cutting delivery time from weeks to a few hours.
Growth Outlook for the Low‑Code AI Segment The low‑code AI solutions market is still consolidating, but adoption dynamics suggest a rapid acceleration: Global adoption: companies with limited budgets, especially in retail, manufacturing, and services, are the first to experiment with these platforms to reduce dependence on specialist talent. Vertical scalability: as training APIs become more efficient, organizations can extend usage to more complex cases, such as fine‑tuning large language models for internal applications. Integrated ecosystems: interoperability with data‑catalog, data‑governance, and CI/CD tools will foster the birth of internal “AI factories,” where the model lifecycle is managed end‑to‑end without manual coding.
💡 Did you know… No‑code AI platforms are already creating a “multiplicative effect”: for every internal data scientist, it is estimated that they can effectively collaborate with up to five non‑technical teams, amplifying the value of AI investment.
In summary, the advent of Trajectory does not eliminate the data scientist role; it transforms it. IT professionals will evolve from model builders to process custodians, while the AI market opens to a broader user base, accelerating global adoption.
---
Follow Me on Social Media
Stay updated on the latest tech news! Follow me on my social channels for exclusive content and tech discussions: LinkedIn - Professional articles and networking Facebook - Community and updates
Don't forget to subscribe to the newsletter to receive articles directly in your inbox!
---
Sources and References Trajectory raises $45M to make AI model training accessible to nontechnical experts Sequoia Capital invests in Trajectory Google Cloud Ventures partners with Trajectory to expand AI accessibility