Open to backend, AI infra & DevOps roles

AnishKumar

Backend & AI Infrastructure Engineer

I build event-driven backend systems and GPU-backed AI model-serving pipelines — FastAPI, Kafka, Kubernetes, on Azure & AWS.

View projects Get in touch
anish@portfolio: ~

$

01 / about

About me

Backend & AI Infrastructure Engineer with production experience building microservice-based, event-driven systems — REST APIs, distributed messaging (Kafka, RabbitMQ), caching (Redis), and containerized deployments on Kubernetes across Azure and AWS.

Strong OOP foundation across Python, Java, and C++, with delivery of scalable backend services, AI/ML inference integration (LLM, computer vision, document intelligence), and production reliability engineering.

B.Tech in Computer Science, BML Munjal University.

  • ~$1MLoop 1.0 annotation platform I built was acquired by a customer
  • 94Azure resources governance-tagged by script, zero failures
  • 6AKS clusters covered by my CrowdStrike Falcon runbooks
  • 57msProduction backend p95 latency, baselined before caching

02 / experience

Where I've shipped

Centific

Associate Application Engineer

Oct 2024 – Aug 2026Chennai, India

01 Backend & AI Engineering

  • Designed and owned microservice-based backend APIs and PostgreSQL-backed services for Loop 1.0, an annotation platform later acquired by a customer in an ~$1M deal.
  • Delivered customer-requested enhancements across REST API behavior, database operations, and service workflows, driving production issue resolution for a live, high-usage platform.
  • Built event-driven microservices with Kafka, RabbitMQ, Redis, and PostgreSQL to decouple ingestion, async processing, caching, and operational storage.
  • Implemented Redis caching to reduce latency and offload primary datastores under production traffic.
  • Containerized services with Docker and deployed on Kubernetes, improving release consistency and reducing manual ops effort.
  • Operated production systems with Prometheus, Grafana, and structured logging to improve reliability and incident response.
  • Integrated AI capabilities into backend services — LLM inference (Llama 70B), computer vision (YOLOv8/v11, NVIDIA VILA & DINO NIM), document intelligence (PaddleOCR).
  • Deployed GPU-backed model-serving workflows using Hugging Face, NVIDIA NGC, Azure, and RunPod.
  • Secured configuration/secrets with JWT, OAuth, Azure Key Vault, HashiCorp Vault.
  • Python
  • FastAPI
  • Kafka
  • RabbitMQ
  • Redis
  • PostgreSQL
  • Docker
  • Kubernetes
  • Azure

02 DevOps & Cloud Infrastructure

  • Automated Azure governance tagging across 94 OneData resources using Azure CLI and the Resource Manager API, enforcing standardized cost-center, project, and environment metadata with zero failures.
  • Diagnosed a production CI/CD defect where deployments pushed new images to Azure Container Registry without updating the App Service image tag, causing successful pipelines to silently leave production on stale code.
  • Designed and deployed Azure Managed Redis for production with high availability, TLS, private endpoints/DNS, monitoring alerts, and an explicit rollback control.
  • Compared Azure Managed Redis against an AKS-based alternative using the Azure Retail Prices API (~$100 vs. ~$605/month), justifying a production-only caching rollout.
  • Implemented Azure Diagnostic Settings and Log Analytics ingestion for production services, then automated telemetry-based cost-estimation reporting within an approved ~$25–63/month budget.
  • Built Azure DevOps CI/CD pipelines for backend and frontend npm package publishing, covering dependency validation, type-checking, linting, builds, and versioned artifact publishing.
  • Baselined production performance (~57ms backend p95 latency), repaired broken frontend Application Insights telemetry, and resolved a noisy-neighbour issue caused by dev workloads sharing the production App Service Plan.
  • Authored reusable CrowdStrike Falcon sensor deployment runbooks (AKS + Linux VM), adopted across 5 Azure subscriptions and 6 AKS clusters.
  • Azure
  • Azure DevOps
  • AKS
  • Docker
  • Azure Monitor
  • Redis
  • CrowdStrike Falcon

Movidu Tech.

Data Analyst Intern

Feb 2024 – Aug 2024Bengaluru, India

  • Built Python ETL workflows (extraction, validation, transformation, structured storage) for large datasets.
  • Developed logistics dashboards and optimized route-planning logic, surfacing operational exceptions for data-driven decisions.
  • Python
  • ETL
  • Dashboards

Adoptev

Web Development Intern

Jun 2022 – Jul 2022Gurugram, India

  • Built responsive ReactJS + Tailwind components and debugged API integrations for cross-device workflow reliability.
  • ReactJS
  • Tailwind CSS

03 / projects

Featured projects

★ Flagship project

Multi-Workspace AI Document Assistant

RAG + LLM tool calling with strict workspace isolation

A full-stack AI web app where users upload documents into separate workspaces and chat with an assistant that answers only from the active workspace's documents. Every answer is cited, it says "I don't know" when the documents don't cover something, and it can take actions through validated tool calls.

  • Python
  • FastAPI
  • React
  • TypeScript
  • PostgreSQL
  • pgvector
  • Gemini
  • Groq
  • RAG
  • LLM Tool Calling
  • Docker
  • GitHub Actions
  • Render
Live demo Demo video GitHub
  1. 01

    Multi-tenant RAG on one shared vector store. Every workspace's chunks live in a single pgvector table, and isolation is enforced inside the SQL vector query, so one workspace can never retrieve another's content. Tests prove it.

  2. 02

    Grounded, cited answers. Hybrid vector + keyword search fused with RRF, a relevance threshold calibrated on real data, server-checked citations, and honest "I don't know" refusals.

  3. 03

    Safe LLM tool calling. An agent loop that saves tasks, posts Discord notifications and runs multi-step searches, with schema-validated arguments, per-message limits, a full audit log, and resistance to prompt injection.

  4. 04

    Reliable by design. Token streaming over SSE, automatic Gemini → Groq model fallback, questions saved before the AI is called so failed answers can be retried, and duplicate uploads ignored.

  5. 05

    Secure accounts. Sign up with email and username, sign in with either. Passwords are hashed with argon2, and sessions use httpOnly JWT cookies.

  6. 06

    Observability. A per-workspace dashboard for latency, token usage, retrieval hit rate and tool success or failure, plus a retrieval-debug view showing exactly which passages an answer used.

  7. 07

    Production-ready. 80+ automated tests and GitHub Actions CI that builds and boots the production Docker image, deployed on Render + Neon entirely on free tiers.

No. 01

ATS Optimize — AI Resume Matchmaking Platform

Full-stack platform that parses resumes and scores them against job descriptions using custom NLP — a recommendation/ranking engine producing keyword-match matrices and compatibility breakdowns. Deterministic AI rewriting module (proper-noun restoration, weakness detection) to improve output without hallucination. Full auth (JWT access/refresh in httpOnly cookies, Google OAuth, email verification) with cross-domain cookie/CORS handling.

  • FastAPI
  • React
  • PostgreSQL
  • JWT
  • OAuth
  • Vite
  • Render
  • Vercel
GitHub
No. 02

AI Document Intelligence Platform

Enterprise-style platform for ingestion, async processing, semantic (vector) search, and AI question answering. Chunking, embedding, metadata extraction, Qdrant vector search, audit logging, RBAC, and Dockerized deployment.

  • FastAPI
  • PostgreSQL
  • Redis
  • Docker
  • Ollama
  • Qdrant
GitHub
No. 03

Distributed Task Processing Platform

Event-driven platform for async workloads with Kafka producers/consumers, retry handling, and fault-tolerant execution. Redis caching, PostgreSQL persistence, and Prometheus/Grafana monitoring with queue metrics.

  • FastAPI
  • Kafka
  • Redis
  • PostgreSQL
  • Docker
  • Grafana
GitHub
No. 04

AI Personalized News Feed: Semantic Ranking & Recommendation Platform

Personalized news platform with RSS ingestion, semantic deduplication (Qdrant), and a transparent multi-signal ranking engine (semantic, freshness, popularity, diversity, interest match). LLM-powered summarization and topic classification with graceful fallbacks, Redis-cached cursor pagination, async Celery workers, React frontend, and full CI/CD.

  • FastAPI
  • PostgreSQL
  • Redis
  • Celery
  • Qdrant
  • Docker
  • React
GitHub

More // on GitHub

  • DNS Lookup CLI

    Python CLI for DNS record lookups, full iterative resolution tracing (root → TLD → authoritative), and CDN-vs-origin latency comparison.

    • Python
    • Networking
    • DNS

  • Web Proxy Server

    Web proxy server built in Python, implementing domain blacklisting and request filtering at the proxy layer.

    • Python
    • Networking
    • Proxy

  • Route Optimization for Logistics

    Optimization algorithms for efficient route planning using Starbucks data, enhancing delivery efficiency and reducing operational costs.

    • Python
    • Jupyter
    • Optimization

  • Drought Prediction

    Machine learning models predicting drought conditions from weather and soil data.

    • Python
    • Jupyter
    • ML

  • SENTANALYS

    Sentiment analysis experiments and models built in Jupyter notebooks.

    • Python
    • Jupyter
    • NLP

  • Endangered.io

    Web browsing website for endangered animals, birds, and sea creatures.

    • HTML
    • CSS
    • JavaScript
View all repositories

04 / skills

My toolbox

  • 01 Languages

    • Python
    • Java
    • SQL
    • C++

  • 02 Backend & APIs

    • Microservices
    • REST API design
    • FastAPI
    • Flask
    • OOP
    • Design patterns

  • 03 Event-Driven & Async

    • Kafka
    • RabbitMQ
    • Redis
    • Azure Event Hub
    • Event-driven architecture

  • 04 Databases & Caching

    • PostgreSQL
    • MySQL
    • Redis
    • Qdrant (vector)

  • 05 Cloud & DevOps

    • AWS
    • Azure
    • Docker
    • Kubernetes
    • AKS
    • Azure Container Registry
    • Azure DevOps
    • Azure CLI
    • ARM API
    • Git
    • CI/CD

  • 06 AI / ML Systems

    • LLM inference
    • Model serving
    • Semantic search
    • Hugging Face
    • NVIDIA NIM

  • 07 Networking

    • VNet
    • Private Endpoints
    • Private DNS
    • VNet Integration

  • 08 Observability

    • Prometheus
    • Grafana
    • Azure Monitor
    • Application Insights
    • Log Analytics
    • Logging
    • Monitoring

  • 09 Security

    • JWT
    • OAuth
    • TLS
    • Azure Key Vault
    • HashiCorp Vault
    • CrowdStrike Falcon
    • Azure App Registrations

05 / education

Education & achievements

B.Tech, Computer Science & Engineering

BML Munjal University, Gurugram, India

2020 – 2024CGPA 8.08/10

Achievements

  • Student Councilor (2023)
  • Sponsorship Lead — 67th Milestone College Festival

06 / contact

Get in touch

Open to senior backend (Python/FastAPI), AI infrastructure and DevOps engineering roles. My inbox is always open — say hello.

anish.26022002@gmail.com
Email me +91 7980737358