Tech Stack
What I Use
A living list of the tools, frameworks, and technologies I rely on to build AI applications and workflows.
Generative AI & LLM Systems
Build practical, verifiable GenAI workflows around OpenAI/vLLM APIs, vector retrieval, tools, and multimodal inputs.
OpenAI / Azure OpenAI & OpenAI-compatible APIs
Integrate cloud OpenAI / Azure OpenAI models and local OpenAI-compatible endpoints (vLLM, LLM Gateway) with structured outputs and tool calling.
On-Prem LLM Serving
Deploy and operate language models within private, security-conscious environments.
vLLM
High-throughput, memory-efficient inference for production LLM serving.
KV Cache & Prefix Caching
Optimize prompt structure and prefix stability to maximize KV cache reuse and reduce repeated prefill overhead.
RAG
Ground model responses in retrievable, source-cited enterprise knowledge.
Vector Databases (ChromaDB)
Index structured technical knowledge with deterministic claim IDs and provenance metadata for source-cited retrieval.
LangGraph & MCP
Orchestrate stateful multi-step agent workflows with LangGraph and expose internal AI pipelines via Model Context Protocol (MCP).
Agent & Tool-Calling Workflows
Connect models with tools, APIs, and structured multi-step task flows.
Multimodal / VLM
Analyze text, images, and document layouts in unified AI workflows.
OCR / ASR
Convert document images and speech into structured, processable content.
Prompt & Context Engineering
Design prompt structures, context compression flows, and output verification checks for reliable, testable AI workflows.
Context Engineering & AI Coding Workflows
Guide AI-assisted development through OpenCode Skills, task decomposition, and repeatable context management patterns.
AI Architecture, LLMOps & Governance
Design secure, compliant enterprise AI systems from feasibility analysis and project leadership to LLMOps deployment.
Enterprise AI Architecture
Shape production AI systems around business, security, and operational constraints.
LLMOps & Production Deployment
Operate internal LLM gateways, model routing, API key governance, usage metering, and offline containerized releases.
AI Security & Data Privacy
Enforce role-scoped read-only access, query isolation from production databases, air-gapped deployment, and audit logging.
Technical Project Leadership & Feasibility Analysis
Partner with cross-functional stakeholders to assess legacy systems, scope adoption strategies, and lead projects end-to-end.
Workflow Orchestration
Coordinate models, data, tools, and human review across end-to-end processes.
API / Tool Integration
Integrate AI capabilities with existing services, systems, and internal tools.
Security & Reliability Design
Build authorization, validation, observability, and failure handling into AI solutions.
Technical Enablement & Knowledge Transfer
Translate platform telemetry and 1-on-1 consultations into reusable self-learning guides and company-wide technical practices.
Backend & Data Engineering
Implement dependable services and data pipelines that make AI workflows usable in production, paired with AI-assisted Web UI delivery.
Python
Primary language for AI applications, automation, data processing, and backend services.
FastAPI
Build typed, high-performance Python APIs for AI and data services.
REST APIs
Design predictable service interfaces for system and tool integration.
SQL
Query, transform, and validate operational and analytical data.
PostgreSQL
Store transactional business state, time-varying org hierarchies, and dual-granularity LLMOps telemetry.
Docker
Package services into consistent, reproducible deployment environments.
ETL / Data Pipelines
Move and transform data reliably across legacy and modern systems.
Databricks
Develop large-scale data engineering, analytics, and machine-learning workflows.
AI-Assisted Web UI Delivery (React / TypeScript)
Core engineering centers on Python/FastAPI and AI systems; steer React and TypeScript interfaces via AI-assisted development to integrate APIs and ship usable internal web apps.
Model Optimization & Applied ML
Optimize inference efficiency and apply predictive modeling with custom objectives to operational problems.
PyTorch
Build sequence neural networks and custom domain-aligned loss functions for operational forecasting.
Time-Series Forecasting
Model temporal patterns to support predictive maintenance and operational planning.
LSTM
Develop sequence models for time-dependent prediction tasks.
Explainable AI
Translate model behavior into evidence engineers and decision-makers can act on.
Optimization
Turn model insights and operational constraints into actionable parameter recommendations.