Anish Shah

Applied AI · New York · 2021—Now

AnishShah

Applied AI builder, educator, and technical storyteller at Weights & Biases

Anish turns emerging AI systems into products, courses, field programs, and public technical artifacts people can inspect and use.

NowHow agents discover tools, use context, and leave evidence teams can trust.

Explore my work
  1. 2021Operational ML
  2. 2022—23Generative systems
  3. 2024—25Evaluation & adoption
  4. NowAgent-native product

Selected technical work

Systems, curricula, writing, and technical sessions.

Representative projects, organized by how the work was built, taught, documented, or delivered.

Systems & Open Source

2024—26

Agent-facing interfaces, evaluation tooling, and ML infrastructure integrations released as public software.

  1. 2026 · Contributor · launch coauthorW&B MCP Server
  2. 2024 · CreatorEvalForge: evaluation-criteria generation & alignment
  3. 2024 · Contributor · three upstream pull requestsW&B logging & checkpointing for Lightning Fabric

Curricula & Workshops

2024—25

Courses and multi-session instruction spanning prompting, LLM evaluation, fine-tuning, and agent engineering.

  1. 2024—25 · Instructor · three coursesAI curriculum: prompting, evaluation & agents
  2. 2024 · Speaker / co-instructor · three sessionsAgentic LLM fine-tuning & evaluation
  3. 2025 · Presenter · two regional broadcasts + replayGoogle Cloud Startup School: GenAI applications

Technical Writing & Implementations

2023—24

Articles and public implementations covering LLM support systems, prompt engineering, and arXiv summarization.

  1. 2023 · CoauthorWandBot: GPT-4 documentation support
  2. 2023 · AuthorPrompt engineering with LangChain & W&B
  3. 2024 · Author · cookbook implementation contributorarXiv summarization with Claude & Weave

Technical Talks & Programs

2023—25

Public sessions on agent architecture, model fine-tuning, and evaluation across conferences and practitioner programs.

  1. 2025 · Speaker / instructor · four public deliveriesArchitecting & orchestrating AI agents
  2. 2023—24 · Speaker · two public deliveriesLLM agent fine-tuning for task automation
  3. 2025 · Conference speakerBuilding agents with evaluation in the loop
Open the complete work index

Evidence beyond the demo

Use. Return. Act.

Independent measures · no shared scale. How measured ↗

  1. Use → return

    Beyond first-run curiosity.

    Mar—Jul 2026
    1. 1K+users · 1K+ organizationscheckpoint Jun 11, 2026
    2. 200K+MCP tool callscheckpoint Jul 29, 2026
    3. More than two-thirdsweekly active users returnedcheckpoint Jun 11, 2026

    Limit Internal analytics; measures are independent.

  2. Learning + action

    Learning coincided with action.

    2024—26
    1. 7K+enrollments · 3 coursescheckpoint Jul 30, 2026
    2. About halfsame-day account associationcheckpoint Jul 30, 2026
    3. 1K+Evals signups · ~36 hourscheckpoint Jan 10, 2025

    Limit Internal analytics; enrollments/signups are not unique learners; association is not conversion.

  3. Programs + pipeline

    Downstream, not just attendance.

    2024—25
    1. $4M+supported event pipelinecheckpoint Oct 1, 2025

    Limit Internal analytics; supported pipeline is not revenue, ARR, or personal causality.