Tech Stack
A practical overview of the languages, frameworks and tools behind my projects. It reflects my GitHub profile README and the technologies used across my portfolio.
Languages
Languages I use or work with, as listed in my GitHub profile.
AI, analytics & systems
Web development
AI & Machine Learning
From model experimentation and VLM evaluation to retrieval and agent workflows.
ML development & experimentation
Training diagnostics, notebook lineage, multimodal experiments and model evaluation.
Retrieval & databases
Document retrieval, chunk auditing, embedding inspection and grounded source evidence.
LLM & multimodal APIs
Typed and streaming API integrations, tool use, structured outputs and error handling.
Applications, Infrastructure & Documentation
The supporting tools for user interfaces, services, reproducible work and technical documentation.
Interfaces
AI demos, component-based frontends, model interfaces and developer tooling.
APIs & deployment
Backend integrations, containerized workflows, version control and open-source releases.
Docs & publishing
Documentation sites, interactive content, reports and template-driven output.
Environment
Operating systems and editor listed in my GitHub README.
What I Build With It
The technical focus behind my open-source work — not just a list of badges.
Evaluation & observability
Experiment diagnostics, deterministic notebook inspection, regression tests, embedding drift, prompt safety and VLM visual-evidence probes.
Retrieval & agent evidence
Source-grounded offline search, chunk quality, context deduplication and tracing of evidence through AI-agent actions.
Local AI & language tooling
CPU-first coding agents, controlled natural-language compilation and explicit safety boundaries around actions.
Token-efficient agents
Receiver-aware context, invariant preservation, quality-constrained token budgets and measurable task success.
Multi-agent & multimodal reasoning
Auditable VLM perceive/describe/critique/refine workflows, occlusion, visual-patch inspection and dataset metadata checks.
Generative audio & creative tooling
Neural audio fundamentals, reproducible acoustic analysis and structured text-to-music prompts.
Engineering Approach
I favor fit to the problem over hype. I choose models, storage and frameworks according to evidence, reliability, complexity and cost — with reproducibility, observability, local-first options and explicit failure modes when the use case requires them.
For concrete examples, see the Portfolio.