Availablefor select projects & roles--:-- IST
Hi, I'm Nikhil 👋
software engineer · backend, cloud & AI systems
I build systems that
scale, ship, and stay up.
Backend services, Terraform-first AWS infrastructure and production RAG, built for founders who need it shipped and teams who need it to stop breaking at 3 a.m.
- shipped at
- RaghuvarTech/
- InkHub / BRMH.ai/
- EatTogether, Seattle
- cert
- AWS SAA-C03
Nikhil Sahuopen to work
Mumbai, India · --:-- IST
−76%AWS bill
<100mspricing quotes
80M+docs indexed
−40%deploy time
// 00 — the shape of what I build
Edge to database to dashboards, all of it observable.
A typical system I design and run. Packets are animated; the stack is real.
prod · ap-south-1 · all greenp95 84ms · err 0.02% · deploy #412
0%
cloud spend cut
Monolith → event-driven containers at InkHub, zero downtime
0%
faster deploy cycles
Environment-gated CI/CD with health-check rollbacks
<0ms
surge-pricing engine
Personalized real-time rates at EatTogether
0M+
judgments indexed
Hybrid RAG with RRF reranking for CaseCloud
// 01 — what I solve
Five problems I get hired for.
Infrastructure, backend, AI and observability — delivered as code you own, with the operational side handled.
Cloud & Infrastructure
Terraform-first AWS environments, GitOps delivery, secrets without long-lived keys, and a path to Kubernetes when you actually need it.
- Terraform / Ansible IaC
- GitHub Actions + OIDC
- Kubernetes, Helm, ArgoCD
- Cost & drift control
$ terraform plan -out=tfplan
+ aws_vpc.main
+ aws_subnet.private[0..2]
~ aws_iam_role.deploy (oidc)
+ aws_db_instance.postgres
- aws_access_key.legacy
Plan: 4 to add, 1 to change, 1 to destroy. ✓ gated
AI, RAG & Agents
Retrieval that cites its sources, agents with guardrails, and LLM features that survive production traffic and budgets.
- Hybrid search + reranking
- Agentic workflows
- Evaluation & cost guards
- Claude / Gemini / open models
query: "limitation period for appeal"
dense · 0.82sparse · 0.77rrf → #1cite [3]
→ "30 days under §…" (3 sources, 1.2s)
Backend & DB engineering
FastAPI, Go and Node services on data models that hold up: event-driven, multi-tenant Postgres, caching and queues, observable from day one.
- REST / GraphQL / gRPC
- Postgres design, RLS & tuning
- Redis, Kafka, event-driven
- Realtime (WebSockets, SSE)
POST/v1/orders20141ms
GET/v1/pricing/quote20018ms
PUT/v1/tenants/acme/users20063ms
GET/v1/fleet/live101ws
POST/v1/events2029ms
POST/v1/orders20141ms
GET/v1/pricing/quote20018ms
PUT/v1/tenants/acme/users20063ms
GET/v1/fleet/live101ws
POST/v1/events2029ms
Observability & SRE
LGTM-stack telemetry, OpenTelemetry tracing across services, alerting that wakes the right person, and incident-ready runbooks.
- Grafana, Loki, Tempo, Prometheus
- OTel tracing & profiling
- SLOs & alert routing
- On-call readiness
p95 latency · apiSLO 99.9% ✓
loki · tempo · promalert → slack #oncall
// 02 — stack, with receipts
Tools I run in production, and where.
Hover a tool to preview how I use it and which shipped systems prove it. Click to pin it while you open the receipts.
daily driver regular working knowledgehover to preview · click to pin · 31 tools · 7 shipped systems
Infra & Cloud
Terraform
Daily driversince 2024 · 2 yrs
Multi-environment IaC: versioned modules, S3 + DynamoDB state, plan-before-apply in CI and nightly drift jobs that open issues.
Receipts
- infra-opsReplaced click-ops and shared credentials with a single IaC repository: reusable Terraform modules, Ansible roles, plan-before-apply pipelines, OIDC auth and nightly drift detection.
- lgtm-sidecarThe full LGTM stack plus profiling and alerting, packaged as a decoupled sidecar that never starves the applications it watches, with distributed tracing across FastAPI and Next.js.
Where
RaghuvarTech
Shipped alongside
// 03 — featured work
Case studies.
// 04 — how I work
Audit first. Ship in increments. Operate it.
01
Audit
Read the code, the cloud bill and the incident history. Name the three things that matter.
02
Architect
Write the design down: boundaries, data flow, failure modes, cost. Review it with you.
03
Build
Small PRs, typed interfaces, tests where they pay. Infra as code from the first commit.
04
Ship
Gated pipelines, canaries where risk is real, rollbacks that are boring.
05
Operate
Dashboards, alerts, runbooks, and a monthly review so the system keeps earning trust.
// 05 — proof of work
Commits, not claims.
Contributions · last 12 months
1,331 total
infra-ops
Centralized multi-environment AWS infrastructure and operations using Terraform, Ansible, and CI/CD (S3 state, DynamoDB locking, SSM config).
HCL 2 0
lgtm-sidecar
Production-ready, resource-optimized LGTM (Loki, Grafana, Tempo, Prometheus) observability sidecar powered by OpenTelemetry (OTel Collector). Provides vendor-neutral distributed tracing, structured logging, and metrics aggregation for Next.js, FastAPI, and Docker microservices via W3C context propagation. SRE/DevOps best practices.
Python 2 0
authverse-mern
Enterprise-grade MERN & TypeScript authentication system featuring JWT session rotation, OAuth 2.0 (Google, GitHub, Facebook), magic link & OTP email verification, and modern React 19 UI.
TypeScript 2 0
Amazon Web Services · 2026-06
AWS Certified Solutions Architect – Associate (SAA-C03)
Verify credentialOpen source
lgtm-sidecar and infra-ops are the same patterns I deploy for clients: resource-capped observability and gated, drift-aware Terraform. Fork them, or let me run them for you.
// 06 — what people say
Don't take my word for it.
Hear it from the teams.
swipe or use the arrows · tap to flip
* illustrative quotes with fictional names; named references available on request.
// 07 — latest writing
Notes from production.
012 min
Advanced RAG Architecture: Graph, Vector, Redis, and PostgreSQL
Combining Graph databases, Vector databases with Reciprocal Rank Fusion (RRF), Redis caching, and PostgreSQL to build a state-of-the-art RAG system.
ragaidatabasesvector db
021 min
The Evolution of AI: From RAGs to Autonomous Agents
Exploring the transition from Retrieval-Augmented Generation (RAG) to fully autonomous AI agents capable of complex reasoning.
aiagentsragllms
032 min
Building a Digital Garden with Next.js and Obsidian
How this site turns an Obsidian vault into a blog, a knowledge graph and an AI knowledge base, and a tour of everything a note can contain.
next.jsobsidiandigital gardenmarkdown
Monthly notes
One email a month.
What I shipped, what broke, and the one idea worth stealing. Also on LinkedIn and Substack.
Read past issues →// 08 — early-stage?
Have an idea?
Rough notes are fine. Tell me what you're building and I'll reply with how I'd architect it, what it would cost to run, and whether I'm the right person.
// 09 — let's talk
Book a 15-min call
Pick a time that fits you, or just
email me at hi@nikhilsahu.dev