/technical career tracks

AI & ML

Craft algorithms that learn, adapt, and mimic the complexity of the human brain as well as the collective intelligence. Push the boundaries of machine cognition by architecting the digital synapses that connect vast information networks.

A digital image featuring a green grid with vertical lines and dots, resembling data visualization or a graph. The background is illuminated with a bright light in the top right corner, creating a glowing effect.
Close-up of a human eye with a vibrant green iris displaying a digital, symmetrical pattern resembling a fractal. The eyes lashes and skin are visible, and the pattern is reflected in the eye, creating a futuristic look.
A close-up view of a microchip surrounded by glowing green lights, resembling a futuristic circuit board. The lights illuminate the chips intricate design, giving it a high-tech, sophisticated appearance.
A close-up illustration of green neurons with interconnecting branches on a vibrant green background, depicting a neural network. The image highlights the intricate structure of nerve cells in a stylized manner.
A glowing green network with interconnected lines and nodes on a grid-like surface, creating a web-like pattern. Its reminiscent of digital or data network schematics.
A digital illustration of interconnected green neural networks with bright nodes against a dark background, symbolizing complex connections and communication pathways.

// 001 // Multimodal AI Engineering // 001 //

// 001 // Multimodal AI Engineering // 001 //

A digital illustration of interconnected green neural networks with bright nodes against a dark background, symbolizing complex connections and communication pathways.
/deep tech career tracks
//Ai & ml

Multimodal AI Engineering

synthesize information across visual, auditory, and textual realms

Key Functions
What are the key areas of focus in this role?
  • Combine and synchronize data from various sources, including text, audio, and video, to create cohesive datasets that AI models can learn from.
  • Design and implement neural network architectures that are optimized for handling multimodal data.
  • Create generative models that produce content which combines text, image, and/or audio.
Problems Solved
What are the types of challenges this role takes on?
  • How can we effectively synchronize and align data from different modalities to ensure our model accurately interprets the combined input?
  • What methods can we use to reduce noise and enhance signal quality across diverse data types to improve our model’s accuracy?
  • How do we design a neural network architecture that can handle the complexities of multimodal data without significant performance trade-offs?
  • What strategies can we implement to enable our models to dynamically prioritize information from different modalities based on context?
  • How can we leverage advancements in one modality, such as natural language processing, to boost performance in another, such as visual recognition?
Overview
What does a typical day look like?

As a Multimodal AI Engineer, you’ll work on technologies that blend audio, visual, and textual data to create AI models with a deeper understanding of the world around them, mirroring human cognitive abilities. The challenges in multimodal AI are complex and varied, and will likely require a deep understanding of different data types and how they can be integrated, as well as experience with machine learning, computer vision, natural language processing, and audio processing. You’ll be at the forefront of developing AI that can see, hear, understand, and interact with the world, creating systems that are greater than the sum of their parts. 

Successful work in this field will require cross-modal data fusion: the development of techniques to effectively integrate and process data from different modalities to create cohesive AI models that can leverage the strengths of each data type for improved understanding and decision-making. You’ll also need to be able to implement synchronized multimodal learning, creating algorithms that enable AI systems to learn from multiple data sources simultaneously, imbuing them with the ability to use insights gained from one modality to enhance the learning and interpretation of other modalities. 

You’ll work on the challenge of multimodal entity resolution — identifying and linking entities such as people, places, and objects, across different data types; for example, connecting a spoken name in an audio clip to a face in a video. Cross-modal retrieval will also be necessary for your systems; they must allow for the retrieval of information across different modalities, such as querying a database with a photo and having it return related text documents. Or vice versa.

The models you build will be capable of some impressive feats: you might build multimodal content generators that can produce content combining text, image, and/or audio; say, generating a video complete with narration and background music, given a written script. You might design sophisticated dialogue systems that can understand and respond appropriately to inputs in various forms, improving human-computer interactions. Through the development of systems capable of scene understanding, you could advance the capabilities of autonomous vehicles, integrating data from sensors, cameras, LIDAR, and radar alongside textual data from maps. Whatever the applications of your work, you will position AI to engage with the world in multidimensional ways.

Roles

Multimodal AI Engineer; ML Engineer - Language & Multimodal Models; Multimodal Research Engineer; Multimodal Language Models AI Engineer; ML Researcher - Multimodal Foundation Models; Research Engineer - Multimodal; Multimodal AI Researcher; Multimodal Generative Modeling Research Engineer.

// 002 // AI & ML Engineering // 002 //

// 002 // AI & ML Engineering // 002 //

A close-up illustration of green neurons with interconnecting branches on a vibrant green background, depicting a neural network. The image highlights the intricate structure of nerve cells in a stylized manner.
/deep tech career tracks
//Ai & ml

AI & ML 
Engineering

Design and optimize algorithms that not only mimic human intelligence but also expand its potential

Key Functions
What are the key areas of focus in this role?
  • Craft and iterate on AI/ML model architectures that are tailored to specific problems, involving a deep understanding of neural networks, decision trees, clustering, and more.
  • Choose the right algorithms based on the specific task and data characteristics, fine-tuning hyperparameters to optimize model accuracy and efficiency.
  • Design models and systems that can scale efficiently and deploy AI models into production environments, integrating with existing technology stacks and ensuring they perform reliably at scale.
Problems Solved
What are the types of challenges this role takes on?
  • How can we design neural network architectures that are more efficient and require less computational power without compromising accuracy?
  • What strategies can be implemented to minimize overfitting while maintaining the model’s ability to generalize well to unseen data?
  • In what ways can quantum computing be integrated into our machine learning models to solve problems currently beyond our capabilities?
  • How can we adapt our learning rate optimization strategy to improve model training efficiency and avoid local minima?
  • Can we develop a system that dynamically adjusts its architecture and parameters in response to feedback from its performance in real-world applications?
Overview
What does a typical day look like?

To be effective as an AI Engineer, you’ll likely need a strong understanding of the mathematical and statistical concepts that power AI algorithms, including linear algebra, optimization techniques, probability theory, information theory, Bayesian statistics, and graph theory, among others. While existing AI/ML libraries such as PyTorch, Keras, and TensorFlow can be used to implement AI models, AI Engineers often need to modify them or even write custom code from the ground up to implement AI solutions in production environments — hence the need to understand how AI algorithms actually work. 

The engineering lifecycle consists of many components; you may specialize in a few of them, or your work might require you to be active throughout the lifecycle. We’ll roughly follow the model development and deployment lifecycle as we outline the types of problems you’re likely to work to solve, keeping in mind that this process is iterative and not necessarily linear.

At the outset, you’ll develop strategies to address imbalanced data, such as oversampling, undersampling, or generating synthetic data so as to improve model fairness and accuracy. You’ll clean, normalize, and transform data into a suitable format for analysis, identifying and/or creating features that improve model performance. You’ll develop novel neural network architectures that push the boundaries of deep learning, improving their ability to learn from complex datasets with fewer parameters or faster training times. You’ll also design adaptive learning rate algorithms that dynamically adjust during model training, achieving faster convergence and improved performance.

Once the data is cleaned, balanced, and preliminary models are built, you’ll employ techniques such as k-fold cross-validation to assess how the results of a statistical analysis will generalize to an independent dataset, crucial for preventing overfitting. Version control and experiment tracking are of utmost importance, and you’ll manage code versions and experiment results meticulously, using tools like Git and MLFlow, to ensure reproducibility and the organized progression of projects. To optimize algorithm efficiency, you’ll focus on computational efficiency, considering that the models may be deployed on devices with limited processing power, such as smartphones or IoT devices. 

The performance of each model must be evaluated: to do so, you’ll implement robust methods to evaluate and compare model performance, using metrics such as accuracy, precision, recall, and F1 score, to ensure that models perform at specified levels. Improving model generalization will also be of concern, and you’ll work on techniques to improve the ability of models to generalize from training data to unseen data, reducing overfitting and making AI systems more robust in real-world applications. You’ll also work to improve transfer learning capabilities, developing learning techniques that allow models trained on one task to be effectively adapted for another. 

Once models are selected for deployment, your final concerns will center on their security and scalability. You’ll assess and enhance the robustness of AI models against adversarial attacks, ensuring their reliability and security in sensitive applications. And you’ll ensure that models can scale efficiently to handle larger volumes of data in production environments, and that they can successfully be integrated with existing technology stacks.

Roles

AI/ML Engineer; AI/ML Software Engineer; ML Engineer - Generative AI; Machine Learning Engineer; Artificial Intelligence Scientist.

// 003 // ML Ops // 003 //

// 003 // ML Ops // 003 //

A close-up view of a microchip surrounded by glowing green lights, resembling a futuristic circuit board. The lights illuminate the chips intricate design, giving it a high-tech, sophisticated appearance.
/deep tech career tracks
//Ai & ml

ML Ops

Operationalize AI, transforming experimental models into robust, production-ready systems

Key Functions
What are the key areas of focus in this role?
  • Automate the end-to-end machine learning lifecycle, from data preprocessing and model training to deployment and monitoring.
  • Use container technologies and orchestration systems to package, deploy, and manage ML models and dependencies across different environments.
  • Develop and implement comprehensive monitoring systems for both model performance and system health by building in alerts for data drift and operational anomalies.
Problems Solved
What are the types of challenges this role takes on?
  • What strategies can we implement to monitor model performance in real-time and detect issues such as model degradation or data drift?
  • How can we automate the retraining process of machine learning models to incorporate new data and maintain their performance over time?
  • What are the best practices for scaling machine learning models to handle increased loads without compromising performance?
  • What is the optimal way to package and deploy machine learning models to ensure compatibility across different environments?
  • How can we leverage feedback loops from production models to inform future model development and refinement?
Overview
What does a typical day look like?

As an ML Ops Engineer, you’ll be responsible for deploying, maintaining, and monitoring machine learning models in production environments, ensuring that they run smoothly, efficiently, and securely. At the intersection of machine learning and operations, you’ll master the full lifecycle of machine learning projects, transforming experimental models into robust applications. Your work ensures that machine learning systems are scalable, reproducible, and maintainable, operationalizing AI applications that evolve with the needs of users and the organization. By implementing monitoring solutions, you’ll keep a vigilant eye on model performance and data drift, maintaining the integrity and reliability of AI applications. 

At a more granular level, your work may consist of some or all of the following: you’ll design and implement automated pipelines that streamline the process of deploying machine learning models into production environments; to ensure that machine learning models remain responsive and reliable as demand fluctuates, you’ll configure and manage infrastructure that efficiently scales to handle varying loads of inference requests; and you’ll develop systems to continuously monitor deployed models for performance degradation and data drift, implementing alerts and triggers for automatic retraining and human intervention as necessary. 

Version control is extremely important for reproducibility and reliability, and to this end you’ll establish robust version control mechanisms for machine learning models and their associated datasets to manage iterations, facilitate rollback, and ensure reproducibility. To verify data quality and consistency, you’ll create processes and tools to validate the quality and consistency of data feeding into production models, preventing issues related to missing data, outliers, or format inconsistencies. Resource allocation will also be a pressing concern, and your work will involve optimizing cloud or other compute resources and infrastructure to balance computational needs with cost constraints, ensuring efficient use of GPUs, CPUs, and memory for the purposes of training and inference. 

Machine learning systems and the data on which they run must be secured against unauthorized access, tampering, and adversarial threats, requiring you to implement security measures to protect machine learning pipelines, models, and data. You’ll need to establish mechanisms for continuous learning and deployment of models, allowing for seamless updates to models with minimal downtime, guaranteeing that models stay current with the introduction of new data. A major challenge will be to reduce latency, and to solve this problem, you’ll identify and address bottlenecks in model inference pipelines to reduce latency, enabling applications that rely on real-time predictions.

Roles

MLOps Engineer; Backend/MLOps Engineer; Machine Learning Scientist; Machine Learning Ops Engineer; Model Serving Engineer.

// 004 // AI Cybersecurity // 004 //

// 004 // AI Cybersecurity // 004 //

A digital image featuring a green grid with vertical lines and dots, resembling data visualization or a graph. The background is illuminated with a bright light in the top right corner, creating a glowing effect.
/deep tech career tracks
//Ai & ml

AI Cybersecurity

obstruct the misuse and infiltration of powerful technologies

Key Functions
What are the key areas of focus in this role?
  • Use an adversarial mindset to anticipate how attackers might exploit weaknesses in AI systems, and develop strategies to prevent such attacks.
  • Protect AI systems from potential threats using threat modeling, risk analysis, and security protocols.
  • Detect and counteract adversarial attacks.
Problems Solved
What are the types of challenges this role takes on?
  • How can we effectively detect and mitigate adversarial attacks on our machine learning models without restricting or compromising their performance?
  • Can we leverage AI to enhance our cybersecurity defenses, and if so, how can we test its efficacy?
  • How do we establish a robust monitoring system to detect any unauthorized attempts to tamper with our AI models?
  • How can we ensure the secure exchange of AI models between partners, preventing IP theft and unauthorized access?
  • In what ways can we automate the process of identifying and responding to security incidents in our AI-driven systems?
Overview
What does a typical day look like?

In this role, you’ll navigate the cutting edge of both cybersecurity and AI, two of the fastest-growing sectors in technology. You’ll have the opportunity to work with advanced machine learning models, understanding their inner workings to better protect them, and deploying AI itself as a tool in your arsenal to predict and neutralize threats more efficiently. Your work will require collaboration with AI developers, Data Scientists, and other Cybersecurity experts, forming an interdisciplinary team dedicated to advancing technology safely and responsibly. 

The dynamic nature of cyber threats, especially in the AI space, means you’ll always be challenged and always be learning. As Ai continues to permeate every aspect of our lives — from healthcare diagnostics to autonomous vehicles, personalized learning, and beyond — the importance of securing these systems from cyber threats cannot be overstated. As an AI Cybersecurity Engineer, you’ll work to prevent AI systems from making incorrect or unsafe decisions by developing defenses that can detect and prevent adversarial attacks.

Your work will consist of identifying AI model vulnerabilities: conducting thorough assessments to uncover weaknesses in AI systems that could be exploited by adversaries, including vulnerabilities in the data pipeline that feeds the model, model architecture, and the deployment environment. You’ll secure data pipelines by implementing encryption, access controls, and monitoring mechanisms to protect sensitive data used in training and operating AI models.

To mitigate adversarial attacks, you’ll develop techniques to detect and counteract adversarial examples (maliciously crafted inputs designed to deceive AI models), to ensure the integrity of model predictions. You’ll also establish systems to continuously monitor AI applications for signs of model drift or data poisoning, ensuring that models remain accurate and reliable over time. 

You’ll establish protocols for the secure sharing of AI models, including digital signing and watermarking, to prevent theft and unauthorized use during model exchange and deployment. To harden AI development environments, you’ll strengthen the security of AI development tools and platforms against vulnerabilities that could lead to code injection, privilege escalation, or unauthorized access to other AI assets.

You might use privacy-preserving techniques such as federated learning of differential privacy to protect user data without compromising model effectiveness. Finally, you could use AI itself as a tool in your arsenal, building machine learning-driven security systems capable of identifying and responding to cyber threats in real-time.

Roles

Principal Cybersecurity Engineer - AI/ML; Cybersecurity Engineer - Cloud & Artificial Intelligence; Senior AI/ML Security Engineer; Security Automation Engineer - AI; Solutions Architect - Cybersecurity AI.

// 005 // Computer Vision Engineering // 005 //

// 005 // Computer Vision Engineering // 005 //

Close-up of a human eye with a vibrant green iris displaying a digital, symmetrical pattern resembling a fractal. The eyes lashes and skin are visible, and the pattern is reflected in the eye, creating a futuristic look.
/deep tech career tracks
//Ai & ml

Computer Vision Engineering

Extend the eyes of technology beyond human limitation

Key Functions
What are the key areas of focus in this role?
  • Design, implement, and optimize computer vision algorithms to solve problems in object detection, image classification, or facial recognition.
  • Use machine learning and deep learning principles such as convolutional neural networks and transfer learning.
  • Process image data for computer vision applications.
Problems Solved
What are the types of challenges this role takes on?
  • How can we enhance the accuracy of our object detection models in varying lighting and weather conditions for autonomous vehicles?
  • What techniques can we employ to reduce the computational complexity of our image classification algorithms without compromising performance?
  • In what ways can we optimize scene reconstruction algorithms to achieve higher precision with fewer input images?
  • What strategies can we use to minimize tracking errors in motion detections systems when dealing with fast-moving or partially obscured objects?
  • Can we apply advanced super-resolution techniques to significantly improve the quality of satellite imagery for environmental monitoring?
Overview
What does a typical day look like?

As a Computer Vision Engineer, you’ll develop algorithms and models that can recognize and/or classify visual data. You’ll work with image and video processing techniques, object detection algorithms, and deep learning tools such as convolutional neural networks to extract and analyze meaningful information from images and videos.

The applications of computer vision are many: you could build systems capable of facial recognition, autonomous vehicle navigation, or medical image classification. You’ll craft algorithms that enable machines to perceive, understand, and interpret the visual world with precision and intelligence akin to, or even surpassing, human perception. 

You’ll develop algorithms that can accurately detect and identify objects within images or video streams, crucial for applications such as surveillance, autonomous driving, and inventory management. Image classification is likely to be one of your main challenges: you’ll create models that can categorize images into predefined classes, useful in sorting visual content, classifying medical images, and more, with a focus on reducing false positives.

You’ll employ 3D geometry and modeling such as stereopsis and computer graphics principles for tasks involving 3D reconstruction, depth estimation, and augmented reality applications. To enable intuitive interactions in gaming, virtual conferences, and smart home devices, you build systems that can interpret human gestures from images or videos. 

Motion detection and tracking is also a common use case for computer vision: you’ll develop systems to detect and track movement in real-time, used in sports analytics, security, and autonomous vehicle applications. You might build models capable of scene reconstruction, rebuilding 3D views from 2D images or video to create detailed models of environments. In the case that the resolution of images needs to be increased beyond the resolution of the capturing device, you’ll apply super-resolution imaging techniques, critical for forensic investigation, medical imaging, and improving low-resolution archives. Another technique that might be frequently pulled from your toolkit is that of semantic segmentation, in which you’ll design algorithms that can understand and segment an image at the pixel level in order to distinguish between different objects and features, vital for applications in autonomous vehicles and robotics.

Roles

Computer Vision Engineer; Software Engineer - ML & Computer Vision; Vision System Engineer; ML Engineer - Computer Vision; Computer Vision Developer; Data Scientist - Computer Vision; Industrial Vision Engineer; Computer Vision Research Engineer; Computer Vision Algorithm Engineer; 3D Computer Vision Engineer.

// 006 // LLM / NLP Engineering // 006 //

// 006 // LLM / NLP Engineering // 006 //

A glowing green network with interconnected lines and nodes on a grid-like surface, creating a web-like pattern. Its reminiscent of digital or data network schematics.
/deep tech career tracks
//Ai & ml

LLM / NLP Engineering

Construct the foundational algorithms that allow machines to understand and generate human language

Key Functions
This is some text inside of a div block.
  • Clear and prepare text data for analysis, using tokenization, stemming, and lemmatization.
  • Extract useful features from text data to convert text into a format that machine learning algorithms can process.
  • Develop algorithms that can analyze text data and respond to natural language inputs.
Problems Solved
This is some text inside of a div block.
  • How can we improve the accuracy of our sentiment analysis model to better capture nuanced emotional expressions in text?
  • How do we effectively incorporate contextual information to improve named entity recognition in diverse domains, such as medical or legal texts?
  • Can we develop a more efficient algorithm for our question-answering system that accurately handles ambiguous queries?
  • What new approaches can we explore to further reduce the error rate in speech recognition under challenging conditions, like noisy environments?
  • How do we measure and improve the semantic text similarity model’s performance to support more effective document clustering and retrieval?
Overview
This is some text inside of a div block.

As an NLP/LLM Engineer, you’ll build systems that can process, interpret, and generate human language. At the confluence of language, technology, and artificial intelligence, you’ll bridge human communication with machine understanding, crafting algorithms that parse human language in ways that machines can process and respond to. You’ll collaborate with interdisciplinary teams of Linguists, Data Scientists, and Software Engineers to develop systems that understand and speak the languages of the world. 

You’ll tackle the complexities of language — its ambiguity, contextuality, and diversity — to teach machines the subtleties of human communication. To accomplish this, you’ll use a variety of techniques, including but not limited to: sentiment analysis — analyzing text to determine the sentiment expressed, whether positive, negative, or neutral, and understanding the emotional tone behind words or phrases; named entity recognition (NER) — identifying and classifying key elements in text into predefined categories such as names of people, organizations, locations, expressions of times, quantities and percentages, etc.; part-of-speech tagging — tagging words in a sentence as nouns, verbs, adjectives, adverbs, etc., which is essential for machines to parse sentence structure. You’ll experiment with cutting-edge technologies like deep learning, transformers, and generative adversarial networks (GANs). 

You will work to preprocess and clean data, as well as augment it when necessary to improve model training effectiveness. You design and implement efficient and scalable neural network architectures tailored specifically for large language models, requiring an understanding of the intricacies of transformer models and attention mechanisms.

Models must be evaluated for their performance, and you’ll design tests and select metrics to accomplish this. It will be helpful to be aware of best practices for deploying and scaling LLMs in production environments, ensuring that they maintain low latency. You may also work on techniques to compress LLMs for efficient deployment in resource-limited environments, enabling a broader application of these models.

There are many use cases for machines that can understand and produce natural human language, and you might find yourself building models capable of machine translation that can provide accurate and contextually appropriate translations across languages. Or, you might build question-answering systems that can understand natural language questions and provide concise, accurate answers by retrieving information from large datasets or documents. You might also build speech recognition systems, enhancing speech-to-text conversion systems for better accuracy and noise resistance, enabling voice-activated applications to understand spoken commands more effectively. Another use for natural language is in text summarization; you might develop algorithms that can automatically generate concise summaries of long texts, such as articles or reports, preserving key information. Or, you might build systems capable of language modeling that can predict the probability distribution of language occurrence, facilitating tasks like auto-completion and text generation.

Roles

NLP Researcher; Applied Scientist - Conversational AI/NLP; Conversational AI Engineer; Natural Language Processing Engineer; Machine Learning Engineer - LLM & GenAI; AI Engineer - NLP/LLM Data; Data Scientist - NLP, LLM, & GenAI; LLM Machine Learning Engineer.

/deep tech career tracks

Additional Career Tracks

Explore other deep tech career tracks