Debashish Ghosal

Getting back into coding after 15 years as a people manager Rest here - https://www.linkedin.com/in/deghosal/

Seattle, WA Joined on Oct 15, 2024 https://github.com/deghosal-2026/

Pinned

Real-world agents break clean evals

Debashish Ghosal

Aug 2

I Built an Agent Eval Harness. Real Agents Broke the Clean Version of the Story

#ai #agentskills #learning #opensource

9 min read

Builder scars reveal untrustworthy agent paths

Debashish Ghosal

Aug 1

Why Agent Evaluation Is Harder Than Model Evaluation

#challenge #ai #agents #programming

8 min read

Want to connect with Debashish Ghosal?

Create an account to connect with Debashish Ghosal. You can also sign in below to proceed if you already have an account.

Create Account

Already have an account? Sign in

Metrics mask silent logic gaps

Debashish Ghosal

Aug 1

Faster PRs, Weaker Instincts: The Judgment Problem in AI-Assisted Engineering

#ai #leadership #llm #agile

4 min read

QA and exception handling get denser

Debashish Ghosal

Jul 31

AI-Assisted Engineering: Faster to Build Isn't Cheaper to Own

#ai #leadership #llm #agentskills

6 min read

Debashish Ghosal

Jul 26

I Planned 10 LLM Evaluation Experiments And Only Ran 1. It Was Enough.

#discuss #ai #webdev #llm

13 min read

Exposes glaring security gaps in multi-MCP setups

Debashish Ghosal

Jul 25

I Connected 3 MCP Servers to One Agent. It Got Scary Fast.

#discuss #ai #mcp #automation

5 min read

Debashish Ghosal

Jul 24

Platform Strategy That Ships: The 90-Day System I Trust More Than AI Roadmaps

7 min read

Overcoming internal dev tool adoption hurdles

Debashish Ghosal

Jul 21

4 Open-Source AI Tools, 1 MCP Server — What I Built and What I Learned

#discuss #ai #python #opensource

15 min read

Debashish Ghosal

Jul 19

The Best AI Productivity Win I Got This Month Wasn't From Coding

#discuss #ai #productivity #softwareengineering

4 min read

Debashish Ghosal

Jul 18

Your Flaky Tests Aren't Flaky

#discuss #ai #python #testing

6 min read

Debashish Ghosal

Jul 17

i built a tool that tracks what AI tasks actually cost. the real number surprised me.

#discuss #ai #llm

6 min read

Debashish Ghosal

Jul 15

i've been building platforms first for 25 years. i think it's wrong now.

#discuss #ai #webdev #softwareengineering

6 min read

Debashish Ghosal

Jul 15

i tested an ai incident commander against 15 real outages — 88% pass rate

#discuss #ai #webdev #python

4 min read

Debashish Ghosal

Jul 12

i spent 3 days on a rag pipeline and the embedding model didn't matter at all

#discuss #ai #webdev #rag

6 min read

Debashish Ghosal

Jul 11

my ai agent ran for 6 hours on a 2-minute task and cost me $200

#discuss #ai #webdev #opensource

6 min read

Debashish Ghosal

Jul 11

i've been doing sdlc for 25 years. ai didn't kill it. it made it the coordination layer.

#discuss #ai #webdev #sdlc

7 min read

Debashish Ghosal

Jul 10

i was using a frontier model to update jira tickets. here's how i fixed it.

#discuss #ai #webdev #llm

6 min read

Debashish Ghosal

Jul 10

i built two langgraph projects and learned that agents are junior engineers with amnesia

#discuss #ai #webdev #langgraph

7 min read

DEV Community

Debashish Ghosal

Badges

2 Week Community Wellness Streak

1 Week Community Wellness Streak

Writing Debut

One Year Club

Currently learning

Pinned

I Built an Agent Eval Harness. Real Agents Broke the Clean Version of the Story

Why Agent Evaluation Is Harder Than Model Evaluation

Want to connect with Debashish Ghosal?

Faster PRs, Weaker Instincts: The Judgment Problem in AI-Assisted Engineering

AI-Assisted Engineering: Faster to Build Isn't Cheaper to Own

I Planned 10 LLM Evaluation Experiments And Only Ran 1. It Was Enough.

I Connected 3 MCP Servers to One Agent. It Got Scary Fast.

Platform Strategy That Ships: The 90-Day System I Trust More Than AI Roadmaps

4 Open-Source AI Tools, 1 MCP Server — What I Built and What I Learned

The Best AI Productivity Win I Got This Month Wasn't From Coding

Your Flaky Tests Aren't Flaky

i built a tool that tracks what AI tasks actually cost. the real number surprised me.

i've been building platforms first for 25 years. i think it's wrong now.

i tested an ai incident commander against 15 real outages — 88% pass rate

i spent 3 days on a rag pipeline and the embedding model didn't matter at all

my ai agent ran for 6 hours on a 2-minute task and cost me $200

i've been doing sdlc for 25 years. ai didn't kill it. it made it the coordination layer.

i was using a frontier model to update jira tickets. here's how i fixed it.

i built two langgraph projects and learned that agents are junior engineers with amnesia