Skip to content
View sadvi11's full-sized avatar

Block or report sadvi11

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
sadvi11/README.md

Sadhvi Sharma

Cloud & Platform Engineer · Calgary, Alberta, Canada

I spent 2.5 years keeping Bell Canada's, T-Mobile US's and Orange's 5G core running against a 99.9% SLA. Now I build on AWS and Azure with the habits that job forces on you.

Open to Cloud · DevOps · Platform · SRE · AI Engineering roles. Calgary · remote across Canada · open to relocation. Permanent Resident — no sponsorship required. Available immediately.

📧 sadhvisharma763@gmail.com · 💼 LinkedIn


Why my background is different

Most cloud engineers learn high availability from a course. I learned it on call, on a telecom core network, where the failure mode isn't a red build — it's someone's call dropping mid-sentence.

At Nokia I operated Cloud-Native 5G Core network functions — AMF, SMF, UPF, CBIS, CBAM, NRF — as containerized workloads on Kubernetes and OpenStack, across deployments for international operators. Zero-downtime rolling upgrades were a contractual requirement, not an aspiration. I delivered a wave-based multi-site 4G→5G migration with zero subscriber downtime. (Nokia appreciation award.)

Then I noticed the thing that made everything since easier: 5G Service-Based Architecture and cloud microservices are the same patterns wearing different names. Service discovery, horizontal scaling, event streaming, container lifecycle.


Three things worth your time

Everything here is public. These three are where the engineering actually is.

Every repository on this profile has a passing CI badge. Not a decoration — several stand up real infrastructure in CI and assert against it, and a few deliberately break their own safety checks to prove the tests can go red. A green suite that has never failed is not evidence.

1. Three-tier e-commerce on AWS EKS — 13 workloads, proven on every commit

8 microservices in 5 languages, 2 databases, a cache and a message broker. Terraform builds the cluster; one templated Helm topology renders all 13 workloads rather than 24 copy-pasted manifests.

What makes it different from the hundreds of other three-tier EKS repos: CI deploys the entire stack to a throwaway cluster on every commit — free, no AWS account — then writes a cart key, deletes the Redis pod, and asserts the key came back. The persistence claim is tested, not stated.

It also took nine CI runs to get green, and the commit history says why: a container running as uid 0 with all capabilities dropped is not root and fails with chown: Operation not permitted; a Java service that looked broken was actually cascading from MySQL being down; and the vendor MySQL image genuinely cannot initialise, which four identical error messages proved.

AWS — EKSAzure — Container Apps
RuntimeEKS — Kubernetes I controlContainer Apps — serverless, scales to zero
IaCTerraformBicep
ScalingHPA on Prometheus metricsHTTP autoscaling, zero to five
Registry authIRSAManaged identity, AcrPull only
Credentials storednone — OIDC federationnone — OIDC federation

What running both actually taught me: identical code, different exposure. /metrics is an in-cluster scrape on EKS and an internet-facing endpoint on Container Apps — the same Prometheus default that is harmless on one cloud publishes your interpreter version on the other. I found that by opening the URL, not by reading the docs.

Try it: https://sentiment-api-6nhl.onrender.com — ask it about financial results, then ask it something in French and watch it return uncertain with its vocabulary coverage rather than guessing.

The Azure deployment is the primary path and is currently offline — that subscription's free-trial credits ran out. Rather than leave a dead link labelled "live", the same service now also runs on a free tier, so the demo survives any one provider's billing state. The Azure pipeline and Bicep templates are unchanged.

Free tier, so it sleeps when idle — the first request takes up to a minute to wake. Measured 42s cold, then instant.

Carrier network functions mapped to AWS service by service — AMF→ALB, CBAM→ECS, NRF→Cloud Map, OAM bus→Kinesis — across 7 Terraform modules with SOC 2 and PCI DSS control mappings.

It's an architecture study with working Terraform, not a production migration, and the README says so. What's real is the source side: I operated these functions, so the mapping is grounded in what they do rather than what their names suggest — including being explicit that the user plane doesn't map cleanly, because a UPF forwarding subscriber packets at line rate is not an ECS task.


Everything else

Infrastructure, delivery and reliability

ProjectOne lineStack
gitops-argocd-kubernetesPull-based delivery with no cluster credential in CI. The pipeline scales prod down by hand and fails unless Argo CD heals it — self-heal confirmed in ~10s every runArgo CD · Kustomize · kind
iac-security-guardrailsCheckov + custom OPA/Rego for Canadian data residency and cost tags. Tested in both directions — which caught a policy that had silently stopped matching anythingOPA · Rego · Checkov · Terraform
sre-incident-practiceBlameless postmortems from my own real incidents, an SLO with an error-budget policy, and the argument for 99.5% over 99.9% because the architecture can't hold the higher numberSRE · SLO · runbooks
azure-finops-guardrailsFinds unattached disks, orphaned IPs, and VMs stopped-but-not-deallocated; forecasts month-end from run rate using median absolute deviation so one spike can't hide inside itAzure · Bicep · Python
azure-devops-pipelinesMulti-stage pipelines with environment gates that live outside the repo, so a developer can't bypass production approval by editing YAMLAzure DevOps · templates
multi-cloud-terraformOne codebase, two clouds, identical network — and the three places AWS and Azure stop being interchangeableTerraform · AWS · Azure
aws-vpc-terraformMulti-tier VPC across two AZs — NAT, tiered security groups, NACLs, least-privilege IAMTerraform · AWS · IAM
flask-ecs-fargate-cicdContainerized API on Fargate, SHA-tagged images, OIDC — no stored AWS keys · live demoECS Fargate · ECR · Actions
docker-flask-ai-appTest-gated CI/CD where the Trivy scan fails the build on HIGH/CRITICAL rather than reporting and moving onDocker · Trivy · ECR
prometheus-monitoring-stackPull-model observability, deliberately isolated from what it watches — a watcher that dies with its host never alertsPrometheus · Grafana

AI and machine learning

ProjectOne lineStack
bedrock-rag-appRAG over financial documents — measured at 195 ms embedding, 225 ms retrieval, ~$0.0003/query. Grounded, so it reports missing context instead of inventing a numberBedrock · Titan V2 · pgvector
smart-ai-agentAgent with tool use, persistent memory and RAG, plus an eval suite for prompt injection and SQL injectionClaude API · pgvector · Flask
structured-test-agentForced tool choice against a strict JSON Schema, with tests asserting the schema stays strict rather than advisoryClaude API · JSON Schema
canadian-financial-sentimentEnd-to-end SageMaker pipeline — train, deploy, serve, tear downSageMaker · S3 · Flask
ai-chatbot-with-memoryCross-session memory in DynamoDB, with a circuit breaker around the model APIFastAPI · DynamoDB
f1-telemetry-pipelineEvent-driven telemetry through a decoupled queue-and-consumer pipeline with DLQ and retrySQS · Lambda · DynamoDB
aws-python-automationboto3 automation — EC2 control, S3, Lambda scheduling, CloudWatch, SNS alertingPython · boto3

What I'd want you to notice

The bugs I found in my own work, because that's harder to fake than a green badge:

  • A REST endpoint that returned a placeholder instead of calling the agent — valid JSON, HTTP 200, and lint and type-checks both passed. Linting proves code is well-formed, not that it does anything.
  • A Rego policy matching nothing while the test suite stayed green, because Terraform emits "tags": null and not r.values.tags is false for null. Fixed the rule, then restructured the tests so one dead policy can't hide behind its neighbours.
  • A green CI badge over a red test suitecontinue-on-error: true with a comment claiming credentials were needed. Eleven of twelve passed offline.
  • A deployed sentiment model scoring 2/6 on real inputs at ~0.52 confidence while cross-validation said 0.471 — worse than random. It now abstains when the input is outside its vocabulary instead of guessing.

A check that cannot fail is worse than no check, because it produces confidence without coverage. Most of my repos ship a WHY.md for the same reason.


What I work with

Cloud — AWS (EKS, ECS Fargate, Lambda, VPC, ECR, S3, DynamoDB, Kinesis, SQS, SNS, IAM, CloudWatch, Bedrock, SageMaker) · Azure (Container Apps, ACR, Bicep, managed identity, Log Analytics, workload identity federation)

Infrastructure as code — Terraform · Bicep · Helm · Kustomize · Docker · Kubernetes · OpenStack

CI/CD & delivery — GitHub Actions · Azure DevOps · Argo CD (GitOps) · OIDC federation

Security & governance — OPA/Rego policy-as-code · Checkov · Trivy · IRSA · least-privilege IAM · SOC 2 / PCI DSS control mapping

Observability & SRE — Prometheus · Grafana · CloudWatch · SLOs and error budgets · blameless postmortems · production on-call

AI/ML — Bedrock · SageMaker · Claude API · RAG · pgvector · agent tool use · LLM evaluation, prompt-injection testing

Languages — Python (boto3, FastAPI, Flask, scikit-learn) · Bash · HCL · SQL

Telecom — Nokia 5G Core (AMF, SMF, UPF, NRF, PCF, UDM) · CBIS/CBAM · CNFs on Kubernetes · 3GPP SBA · ETSI MANO


Everything else, sorted

Twenty-four public repositories is more than anyone will read. This is so you can go straight to whichever one matches the role you are hiring for.

Kubernetes & containers

eks-ecommerce-microservices8 microservices, 5 languages, Terraform + Helm. CI proves the cart survives a pod restart
gitops-argocd-kubernetesArgo CD. CI breaks production on purpose and fails unless it self-heals
mlops-sentiment-eksHPA on Prometheus metrics; CI gates the build on cross-validated model accuracy
flask-ecs-fargate-cicdECS Fargate behind an ALB, no stored credentials. Live demo
docker-flask-ai-appTests gate the build; the image is scanned before it can ship

Infrastructure as code

aws-multi-account-governance6 SCPs, permission boundaries, 31 tests — including the region restriction that silently breaks IAM org-wide
aws-hybrid-network-terraformTransit Gateway. 12 assertions, mostly negatives — prod cannot reach nonprod
nokia-5g-to-aws-migration7 modules mapping carrier 5G network functions onto AWS. SOC 2 + PCI DSS controls
aws-cloudformation-stacksetsOrg-wide baseline. 17 assertions run in both directions
multi-cloud-terraformOne codebase, two clouds — and the three places they stop being interchangeable
aws-vpc-terraformSegmentation modelled on 5G packet core principles. Deployed, screenshotted

Security & governance

iac-security-guardrailsCheckov + custom Rego for Canadian data residency. CI proves the guardrails catch a deliberately insecure fixture
azure-devops-pipelinesWorkload identity federation, Key Vault, approval gates — plus a validator that catches hard-coded credentials
azure-finops-guardrailsFinds resources that cost money and do nothing. Detectors are pure functions, 25 tests, no cloud access

Observability & SRE

sre-incident-practiceReal postmortems from incidents I actually hit. A metrics disclosure, a control that failed open, an OIDC trust mismatch. Open action items left open
prometheus-monitoring-stackPrometheus + Grafana + node_exporter, live dashboards on a real host

AI & data engineering

smart-ai-agentTool use, pgvector memory — graded by a suite including prompt-injection and SQL-injection attacks
bedrock-rag-appTitan Embeddings V2 + Claude Haiku, pgvector RAG. Deployed
azure-sentiment-containerappThe Azure half of the both-clouds comparison. Bicep, scales to zero
canadian-financial-sentimentSageMaker BlazingText, real-time endpoint
f1-telemetry-pipelineEvent-driven SQS → Lambda → DynamoDB → CloudWatch
ai-chatbot-with-memoryConversation state in DynamoDB, async FastAPI
structured-test-agentTool-forced schema output, so the JSON is always valid
aws-python-automationboto3 operations suite — EC2, S3, Lambda, CloudWatch, SNS

Bold ones are where I would start.

Open source

aquasecurity/trivy-checks — contribution accepted to the security scanner used across the industry.


Background

AI Hardware & Technology Specialist — Meta (via Influence Marketing), Calgary · Oct 2025 – present Technical enablement on Meta AI products and XR hardware. Built and delivered training for 50+ retail staff, reducing escalated issues ~30%. The infrastructure work above is built outside working hours.

Solution Engineer, Cloud Core Network — Nokia, Delhi · Dec 2022 – Jul 2025 5G core in carrier production at 99.9% SLA on OpenStack and Kubernetes, for Bell Canada, T-Mobile US and Orange. Wave-based multi-site 4G→5G migration with zero subscriber downtime. Production on-call with structured root cause analysis.

RF & Systems Engineer — AA Electro Magnetic Test Labs, India · 2018 – 2020

Education — M.Tech, Electronics & Communication Engineering, University of Delhi (full merit scholarship) · B.Tech, Shri Mata Vaishno Devi University

Certifications — Microsoft Azure Fundamentals (AZ-900) · AWS Solutions Architect Associate — studying now, sitting September 2026


Get in touch

I'm looking for a Cloud, DevOps, Platform or SRE role where infrastructure is treated as a product rather than a cost centre. If the work above looks relevant to what your team is building, I'd like to hear from you.

📧 sadhvisharma763@gmail.com 💼 linkedin.com/in/sadhvi-sharma-5789a6249 📍 Calgary, Alberta · Permanent Resident, no sponsorship required · open to relocation

Pinned Loading

  1. azure-sentiment-containerappazure-sentiment-containerappPublic

    Sentiment ML API on Azure Container Apps — FastAPI + scikit-learn, containerized, deployed via GitHub Actions CI/CD with Bicep IaC. Serverless, scales to zero. Azure companion to the same model on …

    Python

  2. gitops-argocd-kubernetesgitops-argocd-kubernetesPublic

    Pull-based GitOps with Argo CD — CI stands up a real cluster, breaks production on purpose, and fails unless Argo CD self-heals it. App-of-apps, Kustomize overlays, hardened workloads.

    Makefile

  3. iac-security-guardrailsiac-security-guardrailsPublic

    Policy-as-code that blocks insecure Terraform before merge — Checkov plus custom Rego for Canadian data residency, tagging and IAM. CI proves the guardrails catch a deliberately insecure fixture, n…

    HCL

  4. mlops-sentiment-eksmlops-sentiment-eksPublic

    Production MLOps: sentiment model served with FastAPI on Kubernetes/AWS EKS — HPA autoscaling, Prometheus metrics, Terraform IaC, and CI/CD that gates the build on cross-validated model accuracy.

    Python

  5. nokia-5g-to-aws-migrationnokia-5g-to-aws-migrationPublic

    Nokia 5G Core → AWS architecture mapping: 7 Terraform modules translating carrier-grade network functions (AMF, SMF, UPF, CBAM, NRF, PCF) to production AWS (ALB, ECS Fargate, VPC, Kinesis, DynamoDB…

    HCL

  6. sre-incident-practicesre-incident-practicePublic

    Real postmortems, runbooks and SLOs from incidents I actually hit — a public metrics disclosure, a security control that failed open, and an OIDC trust mismatch. Blameless, with open action items l…

    Python