hiring for ai engineer · /ai
Hiring an AI / Agentic Systems Engineer?
Production RAG, agents with guardrails, and the infra that keeps them cheap.
I build retrieval systems that cite their sources and agents that survive real traffic. On CaseCloud I ground Gemini over 80M+ judgments with hybrid dense+sparse retrieval and RRF reranking; on Immigration Copilot, LLM agents drive a real browser to fill government forms. I care about evaluation, cost guards and observability as much as prompts.
AI EngineerApplied AI / LLM EngineerAgentic Systems EngineerRAG / Search EngineerAI Automation Engineer
80M+
documents in a hybrid-search index
90%
less manual entry via agentic form-filling
RRF
dense + sparse fusion with citations
<2s
p95 grounded answer latency
// 01 — relevant experience
Where I did this before.
- Own infra provisioning across dev/staging/prod with Terraform (VPC, EC2, RDS, IAM) as Infrastructure-as-Code; GitHub Actions pipelines enforce plan-before-apply, drift detection, and SSM/OIDC-backed secrets with zero long-lived credentials.
- Architected infra-ops, a centralized multi-environment IaC platform: S3 remote state, DynamoDB state locking, reusable Terraform modules (networking/compute/storage/IAM) and Ansible roles for PostgreSQL/PostGIS, Nginx+Certbot and PM2 deploys.
- Built lgtm-sidecar, a resource-capped observability stack (Grafana Alloy, Loki, Tempo, Prometheus, Pyroscope, Alertmanager) as a decoupled EC2 sidecar with distributed tracing and rate-capped tail sampling.
- Built an AI-powered BaaS platform generating full-stack e-commerce apps (RESTful APIs, DB schemas, frontend) from prompts, deployed on AWS (EC2, ECS, Lambda, Step Functions, DynamoDB, Valkey, CloudWatch).
- Re-architected monolithic AWS infra into event-driven, containerized microservices, cutting monthly cloud spend by 76% with a zero-downtime migration.
- Owned observability and on-call readiness with CloudWatch alarms and structured logging, reducing mean-time-to-detect and accelerating root-cause resolution.
2025-11 → present
Software Engineer
RaghuvarTech (Raghuvar IT Solutions LLP)
Own infrastructure, CI/CD and observability for a multi-service platform; lead architecture reviews across backend, infra and agentic-AI systems.
2025-06 → 2025-10
Software Development Engineer
InkHub.in / BRMH.ai
Built an AI-powered backend-as-a-service that generates full-stack e-commerce apps from prompts, and re-architected its AWS footprint.
// 02 — selected case studies
Proof, with numbers.
// 03 — skills for this role
The toolbox.
Retrieval & search
- RAG pipelines
- Hybrid search (dense + BM25)
- RRF reranking
- Milvus
- pgvector
- FAISS
- chunking & citations
Agents & orchestration
- Agentic workflows
- Tool use / function calling
- MCP
- LangChain
- Playwright CDP harnesses
- state recovery
Models & platforms
- Claude
- Gemini
- OpenRouter
- prompt caching
- structured outputs
- eval harnesses
Serving
- FastAPI
- streaming (SSE)
- Redis caching
- rate limiting
- cost & token guards
- OpenTelemetry tracing
// 04 — relevant writing
How I think about it.
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
Harnessing Agentic AI for Software Development
How Agentic AI is transforming the software development lifecycle, from automated coding to complex problem solving.
aiagentic aisoftware engineeringautomation
// 05 — recruiter faq
The questions you were going to ask.
What kind of AI work have you shipped to production?
A multi-tenant legal RAG product (CaseCloud) with hybrid retrieval, reranking and citations over 80M+ judgments; an agentic browser harness that fills immigration forms; and a prompt-to-app BaaS that generates full-stack apps. All with cost controls and tracing.
How do you evaluate RAG quality?
Golden question sets per tenant/domain, retrieval metrics (recall@k, MRR) separated from generation metrics (faithfulness, citation accuracy), and regression runs in CI before prompt or index changes ship.
Which models and stacks do you prefer?
Model-agnostic by design: Claude and Gemini in production, open models via OpenRouter for cheap paths. Python/FastAPI for serving, Milvus or pgvector for vectors, Redis for caching, OpenTelemetry for tracing.
Where are you based and can you work remotely?
Based in Mumbai, India (IST). Remote-first; I've shipped with a Seattle team across time zones and I keep 4+ hours of overlap with US Pacific or any EU zone.
What is your notice period?
Currently employed; I can start within 30 days, sooner for contract work.
Full-time or contract?
Open to full-time remote roles and to contract/consulting engagements. Relocation is possible for the right role.
// talk to me about a ai engineer role
Book a 15-min call
Pick a time that fits you, or just
email me at hi@nikhilsahu.dev