Our Technology Stack

We are tool-agnostic but deeply experienced. We choose technologies based on the problem, not the hype cycle. Here is what we work with every day.

LLMs & Foundation Models

Large language models and multimodal AI
GPT-4 / GPT-4o
OpenAI's flagship models for reasoning and generation
Claude 3.5 / Opus
Anthropic models for analysis and long-context tasks
Llama 3
Meta's open-weight models for on-premise deployment
Mistral / Mixtral
Efficient open models for cost-sensitive workloads
Gemini
Google's multimodal models for vision-language tasks

Computer Vision

Detection, segmentation, and image analysis
YOLOv8 / YOLOv9
Real-time object detection and segmentation
PyTorch
Primary deep learning framework for training
OpenCV
Image processing and classical vision algorithms
TensorRT
NVIDIA inference optimization for edge deployment
ONNX Runtime
Cross-platform model inference
Mask R-CNN
Instance segmentation for detailed object analysis

Speech & Audio

Speech-to-text, TTS, and audio processing
Whisper
OpenAI's speech recognition for transcription
Deepgram
Real-time streaming speech-to-text
ElevenLabs
High-quality text-to-speech synthesis
Piper
Open-source TTS for on-premise deployments

Orchestration & Frameworks

Agent frameworks and workflow engines
LangChain
LLM application framework and chains
LangGraph
Stateful multi-agent orchestration
CrewAI
Multi-agent collaboration framework
Celery
Distributed task queues for async processing
Temporal
Durable workflow execution engine

Vector Databases & Search

Embedding storage and semantic retrieval
Pinecone
Managed vector database for production RAG
Weaviate
Open-source vector search with hybrid queries
ChromaDB
Lightweight embedding database for prototyping
pgvector
Vector search as a PostgreSQL extension

Web & Backend

Application frameworks, databases, and APIs
Flask
Python web framework for server-rendered apps
FastAPI
High-performance async API framework
React
Frontend framework for dashboard UIs
PostgreSQL
Primary relational database
Redis
Caching, queues, and session storage
SQLAlchemy
Python ORM and database toolkit

Robotics

Robot operating systems and control
ROS2
Robot Operating System for production robotics
Nav2
Navigation stack for autonomous mobile robots
MoveIt
Motion planning for robotic arms
Gazebo
Physics simulation for testing before deployment
Isaac Sim
NVIDIA simulation for synthetic data and testing

Deployment & Infrastructure

Cloud, containers, and CI/CD
Docker
Containerization for all services
Kubernetes
Container orchestration at scale
AWS
Primary cloud provider (EC2, ECS, Lambda, S3)
Terraform
Infrastructure as code
GitHub Actions
CI/CD pipelines and automation
NVIDIA Jetson
Edge computing for vision and robotics

Observability & Monitoring

Logging, metrics, and tracing
Grafana
Dashboards and visualization
Prometheus
Metrics collection and alerting
LangSmith
LLM tracing and evaluation
Sentry
Error tracking and performance monitoring
PostHog
Product analytics and feature flags

Technology-Agnostic, Not Technology-Indifferent

We have strong opinions on technology selection, loosely held. The right tool depends on your constraints -- latency, cost, compliance, team expertise, and scale. We will recommend what fits your problem, not what is trending on Hacker News.