Ajay Pravin Mahale
@designer-coderajay
About
AI/ML Researcher & Engineer | Mechanistic Interpretability | LLM Safety | NLP | Azure ML & MLOps | PyTorch | TransformerLens
Skills & Technologies
Projects & Repositories
Recent public projects and repositories from this profile.
Circuit-level-interpretability-evidence-does-not-survive-defensible-analytic-variation
PythonCircuit-level interpretability evidence does not survive defensible analytic variation. Pre-registered multiverse over 15,840 specifications; the derived EU AI Act Annex IV claim flips across 73.2% of specification pairs. arXiv:2608.13754
enterprise-agentic-ai-platform
PythonProduction-grade multi-agent AI platform with LangGraph v0.3, MCP servers, hybrid RAG (Qdrant + LlamaIndex), FastAPI WebSocket streaming, and Next.js real-time UI. Deployed on AWS EKS via Terraform.
Portfolio
TypeScriptPortfolio
designer-coderajay
Config files for my GitHub profile.
Causal-Attribution-for-Agentic-Decisions
PythonPost-hoc causal attribution over LLM-agent trajectories. Two total effects that both fail, a closed form for coupling once contexts diverge, and a traceability specification read against Regulation (EU) 2024/1689. Pre-registered experiment published in full and not run.
glassbox-mech
PythonOpen-source EU AI Act Annex IV documentation toolkit. Mechanistic interpretability + circuit discovery for transformers. One function call generates a structured, hash-chained evidence package.