Company overview
Lokos AI — Company overview
Source: https://lokos.ai/
Lokos AI is an AI engineering firm that helps companies identify valuable opportunities for automation and turn them into production systems.
Our team combines machine learning research, software engineering, and operational experience. We work across AI agents, custom applications, models, memory systems, integrations, and analytics—from identifying the right problem through deployment and ongoing improvement.
Our work spans healthcare, financial services, legal research, aviation, media, and consumer technology. Our published observations explore how advances in AI are changing business models, institutions, and everyday life.
AI is no longer just a tool. It will change how your company operates.
Lokos identifies the highest-value pain points in your business and uses the latest AI tools to automate the processes around them. For specialized AI work, we move faster and are more cost-effective than building the same capability through internal hiring.
Industry descriptions and service examples illustrate possible applications. Delivered work is described in Selected work.
Healthcare
Cost efficient agentic care.
Law
Increased access and process automation.
Finance
Mass adoption and direct to consumer tools.
Logistics
A swarm of drones and self-driving vehicles.
Crypto
The currency of thinking machines.
Industrial automation
Closed systems with complete variable control.
Space
Autonomous exploration.
Cannabis
AI-based quality control.
What we build
Eight surfaces. One firm. Owned end to end.
Custom Solutions
Multi-modal · single & cross-modal
Custom AI applications across single and cross-modal domains: audio, text, video, images, and new domains like world models.
Example: Industrial Automation. A multimodal quality-control platform that analyzes camera feeds and sensor data, detects defects, and routes interventions to line operators.
Agents
Multi-layer agentic systems
Fully custom agentic systems built end to end. Multi-layer architectures combining LLMs, retrieval, long-term memory, voice, vision, and tool use into a single deployed product. Text, voice, multimodal, browser, coding, and physical-world control. OpenClaw and NemoClaw as company operating layers.
Example: Healthcare. A voice AI agent that conducts patient check-ins, records structured responses, flags potential risks, and escalates cases to clinical staff.
Models
Classical · deep · generative
Classical and deep learning, supervised through self-supervised, discriminative and generative. We ship predictive analytics, time-series forecasting, world models, embeddings, classifiers, rankers, recommenders, and retrieval pipelines. Diffusion and transformers under the hood, plus whatever replaces them.
Example: Cannabis. A multimodal classification model that combines product images and laboratory data to classify cannabis by chemical profile, product category, and quality grade while flagging contaminants or labeling inconsistencies.
Memory
RAG · vector stores · long-context
Retrieval-augmented generation, vector stores, and long-context memory systems, so models actually know what your company knows.
Example: Law. A permission-aware assistant that searches contracts, policies, and prior matters to answer questions with citations.
Strategy
Roadmaps · build/buy · discovery
AI roadmaps that map opportunities to your data and your goals. We identify where AI compounds, and where it doesn't.
Example: Finance. A practical AI roadmap for applying research agents, automated due diligence, portfolio monitoring, risk analysis, and investor reporting while preserving human approval.
Integration
Forward-deployed engineering
End-to-end deployment into your existing stack with minimal disruption. We embed engineers, ship, and stay until it runs without us.
Example: Crypto. An agent that connects blockchain data, compliance systems, and internal operations to investigate transactions, update records, and route exceptions.
Enablement
Briefings · workshops · transfer
Executive briefings and team workshops on the tools, techniques, and shifts you need to navigate the current environment.
Example: Logistics. A hands-on program that teaches operations teams to supervise forecasting and routing systems, review exceptions, and safely improve automated workflows.
Analytics & visualization
Pipelines · dashboards · reporting
Pipelines, dashboards, and visual reporting that turn your data into decisions.
Example: Space. A mission-control dashboard that combines telemetry, system health, resource consumption, and anomaly alerts to support real-time operational decisions.
The founder: Diego Casabuena
Lokos AI · Est. 2024
Trained at: NYU Center for Data Science
Previously: Podz → Spotify
Countries: VE, ES, UK, IL, CN, US
Hobbies: Martial arts, private aviation
Lokos was founded by Diego Casabuena. He earned a bachelor's degree from NYU with majors in Computer Science and Economics. At NYU, he was a Phi Beta Kappa member and Presidential Honors Scholar. He completed his honors thesis, “Speaking of BERT,” under Rob Fergus and trained as a researcher at the NYU Center for Data Science during the first wave of modern language model work.
Before Lokos, Diego led machine learning at Podz, where his fine-tuned podcast transcription models became the world's first audio feed and helped lead to Podz's acquisition by Spotify. Over three years at Spotify, he progressed to Senior Machine Learning Engineer and shipped podcast previews, the audio-visual translation system that translated Lex Fridman and others, and contributed to Spotify DJ and Auto-Mix.
His work spans speaker diarization, language modeling, clustering, drone calibration, large-scale ML engineering, and economic commentary. Before Lokos, he also co-founded technology companies in the United States and Spain, giving him experience applying technical research inside early-stage businesses.
The firm
A senior AI engineering team that ships production systems.
We embed with your team, build end to end, and turn high-value opportunities into AI products that work in the real world.
Five senior engineers
A focused team with the depth to move quickly.
Embedded delivery
We work inside your team, close to the problem.
End-to-end ownership
Research, data, models, product, and deployment.
Built for production
Reliable AI designed for real users and real workflows.
How we work
Most AI products are vitamins. We build painkillers.
We focus on painful, high-value problems whose solution can create measurable business impact.
Get to production fast. Then iterate with real usage.
We launch, learn, and improve in short cycles using signals from the people who actually use the product.
We build with you, not just for you.
We work alongside your team, transfer knowledge, and leave lasting AI capability inside your company.
Clients
Contact us
Tell us what problem you want to solve.
Email us or book an appointment below. We're based in Lower Manhattan and answer within one business day.
200 Broadway, New York, NY 10038
Legal information
AI systems designed around your business to drive maximum value.
© 2026 Lokos AI · All rights reserved