Skip to content
View AkashKundu114's full-sized avatar
🎯
The system is learning.
🎯
The system is learning.

Highlights

  • Pro

Block or report AkashKundu114

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
AkashKundu114/README.md

Akash Kundu

AI Systems, Autonomous Agents & Machine Learning Research Engineer
Kolkata, West Bengal, India • akashkundu.me • akashkundu7487@gmail.com • LinkedIn


Flagship Systems & Research

  • C.O.P.P.E.R. — Local-First Autonomous Personal AI Operating System

    • 100% offline multi-agent architecture powered by a 14B Sovereign Core fleet across 12 specialized agent types.
    • Sub-millisecond Topological Failure-Predicting (TFP) cascade router (<0.10ms, ~9,856 QPS dispatch).
    • 100% Threat Catch across 350 adversarial test cases via DFM-Guard friction protocol.
    • 685 automated tests (644 Pytest + 41 Vitest) and 25-page IEEE T-CE manuscript.
  • OphthalmoAI — Point-of-Care Retinal Disease Screening Platform

    • Calibrated tri-backbone vision ensemble (DenseNet-201, ConvNeXt-Small, EfficientNet-V2-M).
    • 85.18% 5-class test accuracy, 0.9818 Macro AUROC, and 0.0381 ECE via post-hoc Platt temperature scaling.
    • In-browser edge screening via ONNX Runtime Web WASM (<2MB quantized MobileNetV3-Small INT8, 9.03ms latency).
    • 335 passing tests (312 Pytest + 23 Vitest) with 91.30% DR sensitivity on external IDRiD clinical cohort.
  • Inductive Consensus Clustering (ICC-v2) — High-Dimensional Consensus Clustering

    • Linear O(N·D) unsupervised bimodality feature screening (VSSG) filtering distractor noise with 0.94-0.96 precision@50.
    • Bipartite anchor co-association graph reducing auxiliary memory from O(N^2) to O(NL) (-74.2% peak RAM on D=5,000).
    • Rank #1 cluster recovery on biomedical cancer microarrays (Leukemia ARI = 0.680 vs DEC 0.148).
    • 113/113 automated pytest tests passing; submitted manuscripts to IEEE TKDE and IEEE ICDE.
  • AI-SATHI (Kotha-Khata) — Voice-First AI OS & Financial Ledger

    • Voice-ledger pipeline auto-extracting spoken Bengali into bank-submittable financial PDFs.
    • 2-stage per-chunk grounding RAG chatbot eliminating scheme-matching hallucinations to 0.0%.
    • Sarvam AI + Ollama fallback cascade deployed on Azure Container Apps. Field-tested with 8 rural SHGs.
    • 389 automated tests passing across voice extraction and grounding validation.

Industry Certifications

  • Microsoft Certified: Azure AI Apps and Agents Developer Associate (AI-103)
  • Microsoft Certified: Machine Learning Operations Engineer Associate (AI-300)
  • Oracle Certified: Agentic AI Foundations Associate (1Z0-1157-26)
  • Oracle Certified: AI Database Foundations Associate (1Z0-1195-26)

Core Tech Stack

  • Languages: Python, TypeScript, JavaScript, SQL, C++, Java, Bash
  • AI & Deep Learning: PyTorch, ONNX / ONNX Runtime, Hugging Face, OpenCV, Scikit-learn, CatBoost, NumPy, SciPy
  • Agentic & LLM Systems: LangGraph, Ollama, DeepSeek-R1, Qwen2.5, ChromaDB, RAG, Whisper
  • Backend & Cloud: FastAPI, Node.js, PostgreSQL, SQLite (WAL), Redis, Docker, Azure Container Apps
  • Frontend & Systems: React 19 / 18, Electron, Vite, Tailwind CSS, Three.js (WebGL), PWA (Workbox)
  • Quality & Testing: Pytest, Vitest, Playwright E2E, ESLint, TypeScript Strict Mode, GitHub Actions CI/CD

Verified Engineering Highlights

  • 4 Flagship Systems actively synchronized across repositories with reproducible benchmarks.
  • 1,500+ Automated Tests across production and research codebases.
  • 0.0% Hallucination & 100% Guardrail Intercept in verified safety testing.
  • Live Portfolio & Engineering Showcase: akashkundu.me

Pinned Loading

  1. AkashKundu114.github.io AkashKundu114.github.io Public

    Aura Core — Cognitive Mapping: Personal portfolio & AI engineering showcase for Akash Kundu

    TypeScript 1

  2. OphthalmoAI OphthalmoAI Public

    Point-of-care retinal disease screening prototype: calibrated vision ensemble, Grad-CAM, conformal uncertainty, and offline ONNX edge inference (academic research prototype, non-diagnostic)

    Python 2

  3. AI-SATHI AI-SATHI Public

    Voice-first AI operating system and intelligent financial ledger for rural micro-entrepreneurs & Self-Help Groups (SHGs), powered by Sarvam AI and deployed on Azure.

    Python 1

  4. COPPER COPPER Public

    Sovereign, local-first personal AI desktop OS. Zero-egress multi-agent intelligence powered by Ollama, continuous epistemic decaying memory, proactive Guardian safety firewall, and an Electron HUD.

    Python 2