Skip to content
View nitheshkummarc's full-sized avatar

Organizations

@team-torpedo

Block or report nitheshkummarc

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
nitheshkummarc/README.md

Hi, I'm Nitheshkummar👋

Backend Engineer focused on AI-native systems — where the LLM proposes and deterministic systems decide. I build reliable backend systems, secure retrieval pipelines, and agentic workflows with hard boundaries around what the model is allowed to do.

🚀 Featured Projects

🤖 Revora — Agentic Payment Failure Recovery

The model recommends recovery actions; a deterministic policy engine validates every recommendation before execution. Model output is a suggestion, never an authority. A deterministic resolution layer reconciles late, duplicated, contradictory, and out-of-order webhook events before AI reasoning runs. Execution is only considered successful after system state is verified against the intended outcome.

Tech: Python • FastAPI • Pydantic • Gemini/Groq

📅 PlanPal

Community-driven event-management platform with JWT authentication, role-based authorization, event participation workflows, and a PostgreSQL-backed REST API.

Tech: Python • React • Flask • PostgreSQL • SQLAlchemy • JWT • Tailwind CSS

🔒 Aegis — Enterprise RAG with Transactional RBAC

RAG system where permission checks run inside the same SQL query as vector retrieval, so unauthorized document chunks never reach the LLM. Uses role-aware partial HNSW indexes across three clearance levels, verified against live PostgreSQL EXPLAIN plans. Validated with a 105-test suite and instrumented end-to-end with Langfuse.

Tech: Python • FastAPI • PostgreSQL + pgvector • Next.js • Celery/Redis • Langfuse

📊 Spark Failure Propagation & Root-Cause Analytics

Distributed root-cause analysis platform for Apache Spark — reconstructs execution DAGs from event logs, traces failures through Reverse BFS, and classifies failure scenarios from runtime telemetry with 88.2% accuracy.

Tech: Scala • Apache Spark • PySpark • Hadoop (HDFS/YARN) • Docker

🏗️ WorkLens

Deterministic, explainable candidate-ranking engine that evaluates 100,000 candidate profiles against a job specification in under 2 minutes on a single CPU core, with zero runtime network dependencies.

Tech: Python • Pydantic • Docker

💻 Engineering Focus

Backend Engineering • AI Agent Systems & Guardrails • RBAC / Secure Retrieval • Distributed Systems • Software Architecture • AI Reliability

📫 Connect

📧 Email: nitheshkummarni@gmail.com

💼 LinkedIn: https://linkedin.com/in/nitheshkummar

Pinned Loading

  1. Revora-Agentic-Payment-Failure-Recovery Revora-Agentic-Payment-Failure-Recovery Public

    AI-powered payment failure intelligence that reconstructs payment state, traces root causes, determines safe recovery actions, enforces policy guardrails, and verifies outcomes.

    Python

  2. PlanPal PlanPal Public

    Full-stack event management platform featuring secure authentication, seamless event creation, registration, scheduling, and responsive dashboards for efficient attendee management.

    TypeScript

  3. Aegis-Enterprise-RAG-with-Transactional-RBAC Aegis-Enterprise-RAG-with-Transactional-RBAC Public

    RAG system with RBAC enforced inside the same SQL query as the vector search — unauthorized chunks never reach the LLM. Role-aware partial HNSW indexes across three clearance levels, verified again…

    Python

  4. Spark-Failure-Propagation-Root-Cause-Analytics Spark-Failure-Propagation-Root-Cause-Analytics Public

    Automated root-cause diagnosis for Apache Spark failures using execution DAG reconstruction, dependency-aware Reverse BFS traversal, telemetry feature engineering, and ML-based failure classification.

    Scala 1

  5. WorkLens WorkLens Public

    Deterministic, explainable AI candidate ranking engine, ranking the top 100 candidates from 100K profiles using capability, behavioral, and honeypot-aware scoring.

    Python