Skip to content
View prithidevghosh's full-sized avatar
  • India

Block or report prithidevghosh

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
prithidevghosh/README.md

Hi 👋, I'm Prithidev

Backend Engineer · Shipping production LLM systems end-to-end · Founder of Askube

Top 100 of 31,000 teams

profile viewstwitter

🚀 What I'm building

Askube — GenAI content-repurposing SaaS, live in production, built solo

  • One video → LinkedIn posts, X threads, auto-captioned clips & newsletters in the creator's brand voice, in ~58s
  • Full pipeline end-to-end: transcription, LLM generation, clip detection, video chat (RAG), subscription billing, publishing to 5+ platforms
  • PythonFastAPILLM pipelineDocker

blackbox — fully local flight recorder for dev work npm

  • Records terminal commands, AI-agent sessions & git commits into local semantic memory — ask "what broke the deploy" in plain English, get cited answers; drafts standups & RCAs with a local LLM
  • ⚡ flashback: when a failure repeats, the fix that worked last time appears right under the failed command
  • 100% local & private (secrets redacted before ingestion) · 50 unit tests + autonomous e2e harness · npm releases via OIDC trusted-publishing CI with SLSA provenance
  • Node.jszsh hooksSupermemory LocalOllama

MindFlayer — RL environment for emergent LLM deception 🏆 Top 100 / 31,000 teams

  • A trainable LLM agent (Qwen2.5-0.5B) learns to deceive three fixed GPT-4o-mini investigators, scored by an LLM judge on a Theory-of-Mind taxonomy
  • SFT warmup + 500-step GRPO · ToM score 0.041 → 0.103 · anti-reward-hacking penalties converged to zero
  • Live on Hugging Face Spaces · PyTorchGRPOTRLOpenEnv

Kneedle — on-device health AI for osteoarthritis patients

  • Offline-first gait analysis + Gemma-powered conversational AI (Google Gemma 4 Good Hackathon)
  • Gemmaon-device inference

🛠️ Languages and Tools

pythonpytorchfastapijavascriptnodejsexpressredismongodbmysqlelasticsearchdockerkubernetes

📊 Stats

top languages

streak stats

📫 Reach me

prithidevghosh@gmail.com · LinkedIn · X/Twitter · getaskube.com

Pinned Loading

  1. mindflayermindflayerPublic

    MindFlayer — Reinforcement Learning Environment for Emergent Deceptive Behavior in LLM Agents

    Python 1

  2. kneedle-appkneedle-appPublic

    Dart

  3. kneedle-backendkneedle-backendPublic

    Python

  4. kneedle-landingkneedle-landingPublic

    TypeScript

  5. queryforgequeryforgePublic

    SQL Debugging & Optimisation Environment

    Python