Skip to content
View Ashut0sh-mishra's full-sized avatar
4 followers · 8 following

Achievements

Achievement: YOLOAchievement: Pair Extraordinairex2Achievement: Pull Sharkx2Achievement: Quickdraw

Achievements

Achievement: YOLOAchievement: Pair Extraordinairex2Achievement: Pull Sharkx2Achievement: Quickdraw

Block or report Ashut0sh-mishra

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.

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
Ashut0sh-mishra/README.md
Ashutosh Mishra — Software Engineer

Software Engineer | Java · Spring Boot · Keycloak/IAM · AI

NIELIT, MeitY (Govt. of India) · ~3 years · Java · Spring Boot · Keycloak · Oracle 19c · FastAPI · PyTorch

NIELIT OSS IAM Status

📄 Resume / CV

📊 Numbers first
1.3M+ Indian Railways employees on the IPAS payroll platform I helped modernise at CRIS (Java 17 · Spring Boot · Oracle 19c)
25K+ Users on a unified Keycloak SSO layer across 7 production platforms at NIELIT (OAuth2 · OIDC · JWT · RBAC)
96.1% Crop disease detection accuracy across 21 classes — AgriAnalyze (YOLOv8, 75K images)
~70% Oracle SQL query speedup via execution plan analysis and index redesign — PayNways fintech

🔑 What makes this profile different

  • Not a tutorial-project portfolio — every project here runs at production scale or is published research
  • Keycloak/IAM experience is genuinely rare — I've implemented SSO federation, realm configuration, and RBAC policy from scratch for a 25K-user multi-platform environment
  • OSS contributions to a Michigan State University research lab's core framework — constraint optimizer architecture, tensor ops, and inference engine improvements, not documentation fixes
  • Enterprise Java at scale — Oracle 19c, JBoss EAP, IBM WebSphere, Spring Boot on platforms where uptime means actual employee salaries

💼 Professional Experience

NIELIT — National Institute of Electronics and Information Technology (MeitY, Govt. of India)

AI Platform & Infrastructure Engineer

  • Designed and deployed enterprise Keycloak SSO across 7 production platforms: OpenEdX, JupyterHub, BigBlueButton, Indico, Guacamole, Apache Superset, WordPress
  • Implemented OAuth2 / OpenID Connect / JWT / RBAC — single sign-on for 25K+ users across the EChipHub platform
  • Managed infrastructure on AWS EC2 with Docker-based deployments; integrated Apache Superset + ClickHouse analytics stack
  • Keycloak OAuth2 OIDC JWT RBAC Docker AWS EC2 Spring Boot Spring Security

CRIS — Centre for Railway Information Systems (Deputation via NIELIT)

Backend Engineer

  • Contributed to modernising IPAS — the Indian Railways integrated payroll platform serving 1.3M+ employees
  • Backend development using Java 17, Spring Boot, and Oracle 19c; developed REST APIs and enterprise web application modules
  • Deployed and maintained on JBoss EAP and IBM WebSphere application servers
  • Optimised Oracle SQL queries for payroll processing workloads; production issue resolution and support
  • Java 17 Spring Boot Oracle 19c REST APIs JBoss EAP IBM WebSphere Oracle SQL Spring Security

PayNways — Fintech Startup

Backend Engineer

  • Backend development and production support for fintech payment workflows
  • Migrated codebase from Java 8 to Java 17; implemented PingFederate SSO integration
  • Oracle SQL optimisation — ~70% query speedup via execution plan analysis and index redesign
  • Java 8/17 Spring Boot PingFederate SSO Oracle SQL REST APIs

🚀 Featured Builds

🌾 AgriAnalyze — AI Crop Disease Detection

Production computer vision pipeline for Indian wheat and rice field disease identification.

  • YOLOv8 · 1.4M params · 75K-image HuggingFace dataset · 21 disease classes · 96.1% accuracy
  • Three-config ablation study · McNemar test + Holm-Bonferroni correction · EML cost model
  • YOLOv8 FastAPI React HuggingFace Computer Vision PyTorch
  • 🔗 Live Demo | Repo

🤖 NEXUS AI — Autonomous Slide Generation Platform

Production-style autonomous multi-agent platform for AI-powered presentation generation.

  • 6-step agent pipeline: analyse → search → plan → generate → assemble → export
  • Real-time SSE progress streaming · Celery + Redis async workers · PDF and PPTX export via Cloudflare R2
  • FastAPI Celery Redis Claude AI React Playwright Docker

🔋 Safe RL for BESS — Reproducible Benchmark

Diagnostic benchmark for battery energy storage control — targeting NeurIPS Safe RL Workshop.

  • PPO · PenaltyPPO · LagrangianPPO · ProjectionPPO · MPC oracle baseline
  • git commit hash + seed locking · 200K-step training budget · 25 evaluation episodes per method
  • PyTorch Safe RL Gymnasium Reproducible ML

🔐 EChipHub SSO Platform (NIELIT)

Unified identity layer across 7 production EdTech and research platforms on AWS.

  • OpenEdX · JupyterHub · BigBlueButton · Indico · Guacamole · Apache Superset · WordPress
  • OAuth2 / OIDC / JWT / RBAC · 25K+ active users · Single sign-on, zero password sprawl
  • Keycloak Docker AWS EC2 Spring Boot OAuth2 OIDC Spring Security

🛠 Rare skill combination

Keycloak / IAM       OAuth2 · OIDC · JWT · RBAC · Realm federation · SSO · PingFederate
Enterprise Java      Java 8/17 · Spring Boot · Spring Security · Microservices · Oracle 19c · JBoss EAP · IBM WebSphere
Production AI        PyTorch · YOLOv8 · FastAPI · Celery · Claude AI · Computer Vision
Research OSS         Neuro-symbolic AI · Constraint optimisation · Tensor ops · Inference engine

🤝 Open Source — DomiKnowS, MSU HLR Lab · 10 Merged PRs

Core framework contributions to DomiKnowS — a neuro-symbolic AI research framework from Michigan State University's HLR Lab. These are internal framework changes (constraint optimizer architecture, inference engine improvements, GPU-compatible tensor operations, developer tooling) reviewed and merged by MSU researchers.

PR Change
#442 fix(eqL): auto-wrap scalar values and forward kwargs
#440 Make constraint optimizer pluggable
#439 feat(lossprogram): add training_style + Gumbel hyperparameters to InferenceProgram
#438 fix(dataNode): explicit root selection instead of ordering-dependent logic
#432 fix: expand nested LC results during variable expansion
#431 Fix #358: Return NaN for ifSatisfied when no antecedent is True
#430 Add iffL (if-and-only-if) biconditional constraint support
#429 fix(lcUtils): improve error reporting for undefined variables
#428 feat(dataNode): handle multiple roots with isRootUnique and addBatchRootDN
#427 fix(metric): use tensor operations when calculating metrics

Other contributions: PostHog (YC W20) · AllenAI scispaCy · Microsoft FarmVibes.AI

📬 Find Me

LinkedIn GitHub X Resume

Scan to connect instantly

LinkedIn GitHub
LinkedIn QR GitHub QR
Enterprise Java · Keycloak IAM · Production AI · Statistically validated research.

Pinned Loading

  1. agri-analyze agri-analyze Public

    AI-powered crop disease detection system

    Python 1 2

  2. DomiKnowS DomiKnowS Public

    Forked from HLR/DomiKnowS

    Python

  3. nexus-ai nexus-ai Public

    Python

  4. posthog-js posthog-js Public

    Forked from PostHog/posthog-js

    Send usage data from your web app or site to PostHog, with autocapture.

    TypeScript