AWS Cloud Build & Advisory Engagement
Client: Arham Technosoft

The Challenge

Arham Technosoft’s corporate gifting business ran on large, spreadsheet-driven catalogues — hundreds of products spread across dozens of categories. Clients described what they wanted in plain language, such as “premium gifts under ₹750 for Doctors’ Day,” but matching those requests by scrolling a spreadsheet was slow and quietly missed good options.

• Natural-language client requests could not be matched by keyword search alone
• Budgets stated in a client’s request had to be honoured exactly across every suggestion and multi-item hamper
• Every enquiry needed to become a consistent, professional, client-ready quote
• Arham Technosoft wanted to move at the speed of a client conversation — without building or operating any
AI infrastructure of its own

The primary goal was a fast, accurate, on-budget quoting workflow, delivered without the overhead of hosting or
managing AI models.

Discovery

DevOps TechLab ran a structured discovery phase with Arham Technosoft’s technical and business stakeholders to translate the gifting workflow into a concrete technical requirement set.

• Assessed the catalogue scale: 854 products across 12 vendors, requiring both semantic (text) and visual
(image) search
• Evaluated build-your-own vs. managed-AI approaches, and selected Amazon Bedrock to avoid hosting or
fine-tuning foundation models directly
• Defined the core product requirements: natural-language search, photo-based search, budget-aware hamper
building, and branded PDF quote export
• Captured non-negotiable security requirements: no publicly reachable database, no third-party AI API keys,
and IAM-only access to AI services

This phase established a clear technical baseline and a shortlist of AWS-native services capable of meeting both the product and security requirements.

Onboarding

Based on discovery findings, DevOps TechLab designed a fully serverless architecture and validated it with Arham Technosoft before implementation began.

• Designed the platform around AWS Lambda, Amazon API Gateway, and Aurora Serverless v2 with the
pgvector extension, keeping relational and vector data in a single database
• Selected Amazon Bedrock foundation models — Titan Text Embeddings V2, Titan Multimodal Embeddings
G1, and Amazon Nova Lite — for all generative AI functionality
• Reviewed the architecture and security model with Arham Technosoft’s stakeholders before build began
• Planned a phased delivery to de-risk rollout of a production-facing internal tool

This onboarding step ensured the architecture, cost model, and security posture were agreed before any
infrastructure was deployed

Operations & Support

DevOps TechLab built, deployed, and hardened the platform end to end inside Arham Technosoft’s own AWS
account (ap-south-1 — Mumbai), with continuous validation at each stage.

• Built natural-language and photo-based product search, with budget stated in a query enforced as a hard filter
• Implemented the hamper builder — one product per category, no duplicates, ranked to use the stated budget
well rather than under-spend it
• Embedded all 854 catalogue products using Titan embeddings, indexed with HNSW in Aurora PostgreSQL
for fast nearest-neighbour search
• Delivered branded, client-ready PDF quote export from finished hampers
• Hardened the environment: secrets fetched from AWS Secrets Manager at runtime, a private database
reachable only via Amazon RDS Proxy and an SSM bastion, Origin-Access-Control-protected S3 buckets,
and VPC endpoints in place of a NAT gateway
• Deployed the full stack via AWS CloudFormation, so the same build is reproducible across accounts

All changes were rolled out with validation at each step, ensuring a stable, production-ready platform with no public database and no AI keys anywhere in the codebase.

Optimisation & Advisory

After launch, DevOps TechLab focused on keeping running costs low and predictable, and on advising Arham
Technosoft’s roadmap for hardening the platform further.

• Right-sized Aurora Serverless v2 to scale to its floor when idle, keeping the database as the largest but still bounded cost driver
• Replaced a NAT gateway (~$35/month) with VPC interface and gateway endpoints, cutting a recurring fixed
cost
• Delivered a predictable total running cost of approximately USD 90–130 per month for the full environment
• Recommended a forward security roadmap: scheduled rotation of database and administrative credentials,
migration to Amazon Cognito for per-user authentication and audit trails, and custom domains with managed
certificates

This advisory work ensures Arham Technosoft has a clear, low-effort path to harden and scale EasyGift as usage
grows.

Summary

By combining Amazon Bedrock, AWS Lambda, Aurora Serverless v2, and a fully serverless design, DevOps
TechLab helped Arham Technosoft turn corporate gifting from a manual catalogue exercise into an AI-assisted
workflow. A client’s request — typed in plain language or shown as a photograph — now becomes ranked products,a budget-compliant hamper, and a branded PDF quote in minutes, running at roughly USD 90–130 per month with no AI infrastructure to manage and no database ever exposed to the public internet.

About DevOps TechLab

DevOps TechLab helps organisations design, build, and operate cloud-native and AI-powered platforms on AWS combining serverless architecture, managed AI services, and strong security practices aligned with the AWS WellArchitected Framework.