Skip to content
View SakshamDevloper's full-sized avatar
🎯
Focusing
🎯
Focusing
17 followers · 11 following

Achievements

Achievement: YOLO

Achievements

Achievement: YOLO

Block or report SakshamDevloper

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
SakshamDevloper/README.md
terminal boot

Saksham Sethi

AI/ML Engineer · GenAI & Multi-Agent Systems · Full-Stack Backend 




┌─ session log ──────────────────────────────────────────────────────────┐
│ user      saksham                                                      │
│ role      B.Tech IT (final year) → AI/ML Engineer                      │
│ base      Jammu, India                                                 │
│ uptime    building since 2023                                          │
│ mode      research + shipping, alternating                             │
│ status    open to internships · full-time · remote                     │
└──────────────────────────────────────────────────────────────────────┘

research_interests[]

I care less about wrapping an API call and calling it "AI," and more about the boring, unglamorous part — state machines, evaluation harnesses, fallback routing, the stuff that decides whether a demo survives contact with real input.


experiments/ — active & completed

idhypothesismethodstatus
exp‑01 An explicit multi-agent state machine outperforms a single mega-prompt on tool-heavy tasks 4-role LangGraph pipeline (Planner/Researcher/Executor/Critic) + fine-tuned Mistral-7B via vLLM 🟡 in review
exp‑02 A knowledge graph beats flat vector search for messy industrial documents (PDFs, P&IDs, scanned logs) Agentic RAG over Neo4j, built for ET AI Hackathon 2026 🟢 shipped
exp‑03 A CNN can hit clinically-relevant accuracy on tumor/normal MRI classification with a modest dataset TensorFlow/Keras CNN + transfer learning, full eval pipeline 🟢 97% test accuracy
exp‑04 Adding gradient boosting on top of ARIMA residuals improves retail forecast error ARIMA(1,1,1) baseline → ARIMA + XGBoost + feature engineering 🟢 MAPE 13% → 12.98%
exp‑05 A voice-first interface reduces friction for everyday desktop tasks more than a chat window does Wake-word detection, neural TTS, system automation (ATHENA) 🟢 active daily driver

selected_work/

Conductor — repo

An enterprise multi-agent platform built as an explicit LangGraph state machine instead of a black box — four agents hand off work with visible, debuggable transitions.

Python LangGraph FastAPI Next.js vLLM

KnowledgeBrainAPI — repo

Agentic RAG that turns a pile of unstructured industrial documents into a knowledge graph you can query in plain language.

Python Neo4j RAG

Nexus / ATHENA — repo

A desktop voice assistant — wake-word detection, neural TTS, live automation — built to actually be used, not demoed once.

Groq LLaMA 3.3 Azure TTS

Brain Tumor Detection — repo

CNN classifier for MRI scans, 97% test accuracy, with the full train/eval pipeline documented rather than just the headline number.

TensorFlow OpenCV
ls more_repos/
repo stack note
Hostel Management System MERN, Socket.io real-time, role-based, JWT auth
Sales-Forecasting-x-ML Statsmodels, XGBoost second-iteration forecasting
RAG-Powered-AI-Assistant LangChain source-grounded retrieval
Voice-Enabled-Task-Management-System TypeScript voice + manual task manager
Full-Stack-Portfolio MERN 12+ demo apps

metrics/ — self-updating, refreshed daily via GitHub Actions

isometric commit calendar language breakdown achievements

⚙️ These three panels are generated by a scheduled workflow (metrics.yml, included alongside this README) — they redraw themselves every day rather than re-querying an API on every page load like a typical badge does. Setup is in the note at the bottom of this message.


leetcode/

LeetCode Stats


contributions/

snake

credentials/


last deploy: auto · this file is versioned like everything else I ship

Pinned Loading

  1. LangGraph-Multi-Agent-Orchestration-System LangGraph-Multi-Agent-Orchestration-System Public

    Conductor — enterprise multi-agent AI orchestration platform. Four agents (Planner, Researcher, Tool Executor, Critic) run as an explicit LangGraph state machine over your documents and live tools …

    Python 1 1

  2. Nexus-AI-Assistant Nexus-AI-Assistant Public

    ATHENA is an AI-powered desktop voice assistant with real-time speech recognition, LLM integration, and system automation. It performs tasks like opening apps, setting reminders, fetching weather, …

    JavaScript 2

  3. Full-Stack-Portfolio Full-Stack-Portfolio Public

    Complete development portfolio featuring 12+ diverse projects. Demonstrates proficiency in MERN stack (MongoDB, Express, React, Node.js), RESTful APIs, full-stack web development, and modern JavaSc…

    JavaScript 1

  4. Brain-Tumor-Detection Brain-Tumor-Detection Public

    AI-powered brain tumor detection system for medical imaging analysis. Leverages convolutional neural networks to classify MRI scans as tumor-positive or normal. Includes curated dataset with train/…

    Jupyter Notebook 1

  5. HostelManagementSystem HostelManagementSystem Public

    MERN-stack hostel management platform with real-time dashboard, role-based access, and Socket.io live updates for rooms, payments, complaints, and more.

    JavaScript 1

  6. KnowledgeBrainAPI KnowledgeBrainAPI Public

    RAG-powered, agentic AI platform that ingests industrial documents (PDFs, P&IDs, scanned forms, maintenance logs) into a unified knowledge graph — queryable in natural language in seconds. Built fo…

    Python 1