Token Meter
A local-first dashboard for what coding agents spend: tokens, context, cost, and time. Works with Claude Code, Codex, Cursor, OpenCode, Kiro, and Pi.
Pratik BhavsarAI Engineer · Cisco
I ship AI across the stack.
Developer tools, agent skills, open models, public benchmarks, and the books that explain them.
Jina AI → Enterpret → Galileo → Cisco
25K followers on LinkedIn
3K followers on X
Featured in Andrew Ng's The Batch
Guest on Latent Space
Top AI Developer to watch, 2023
M.Tech IIT Bombay
Tools
Tools that show what AI work costs.
Skill packages
Skills that teach coding agents to debug AI apps.
Eval Engineer
A skill package that turns production traces into a root-cause diagnosis, a bounded fix plan, and a verification plan.
Models
Open models for grounded answers.
BRAG
Qwen2 and Llama 3 models fine-tuned for retrieval-augmented generation, built to answer from evidence rather than memory.
- BRAG-Qwen2-7B7B · 128K context53.23
- BRAG-Llama-3.1-8B8B · 128K context52.29
- BRAG-Llama-3-8B8B · 8K context51.70
- BRAG-Qwen2-1.5B1.5B · 32K context46.43
ChatRAG-Bench score
Benchmarks
Public benchmarks for how models really behave.
Agent Leaderboard v2
30+ models tested on 500 multi-turn support scenarios across banking, healthcare, investment, telecom, and insurance.
Books
Five field guides on building production AI.
Eval Engineering
How to build trust in production AI: metrics, LLM judges, expert review, small eval models, and guardrails.
Writing
Pakodas, a newsletter on becoming a better AI engineer.
Choosing model vs harness
The model is the engine; the harness is the car.
AI engineering
Evals
Talks
Eleven talks on evals, search, and ML systems.
Experience
Early engineer, then builder in the open.
I take AI from research to shipped product. I also founded Maxpool, a community for AI engineers, and write Pakodas.
- 2026 – nowCisco
Tokenomics and evaluation for AI agents.
- 2023 – 2026Galileo
Open-source evals and benchmarks, plus five books. Acquired by Cisco.
- 2021 – 2023Enterpret
Founding engineer. Semantic search, NLP pipelines, and MLOps.
- 2020Jina AI
Employee #6. Open-source neural search.
- 2017 – 2020TaskHuman · Morningstar
Search, recommendations, and quantitative ML.