Executive summary
Arham Technosoft runs a corporate gifting business with large, spreadsheet-driven catalogues: 854 products from 12 vendors. Clients ask in plain language, keyword search missed good options, and every enquiry needed an exact-budget, client-ready quote. DevOps TechLab built EasyGift, a fully serverless AI platform on Amazon Bedrock, that searches by words or by photo, builds budget-true hampers and exports branded PDF quotes in minutes, without any AI infrastructure to manage.
- Client
- Arham Technosoft · Corporate gifting
- Business problem
- Large catalogues, plain-language requests and exact budgets made every quote slow and inconsistent.
- What we delivered
- EasyGift: AI search by words or photo, a hamper builder and branded PDF quotes, on Amazon Bedrock.
- Scale
- 854 products from 12 vendors, embedded for search
- Running cost
- About US$90–130 a month for the full environment
The business challenge
Arham Technosoft’s catalogues ran to hundreds of products across dozens of categories, held in spreadsheets. Clients asked in their own words, such as “premium gifts under ₹750 for Doctors’ Day”, and every suggestion and multi-item hamper had to stay within budget.
Why it mattered
- Keyword search missed good options
Plain-language requests rarely matched catalogue wording, so good products were overlooked.
- Budgets had to be exact
Every suggestion and every multi-item hamper had to honour the client’s budget precisely.
- Every quote built by hand
Each enquiry needed a consistent, client-ready quote, assembled manually.
- No appetite for AI infrastructure
The business wanted AI without hosting models or running AI infrastructure in-house.
What Arham Technosoft set out to achieve
- Search the full catalogue by meaning, in words or with a photo.
- Treat the budget as a hard limit on every suggestion and hamper.
- Build complete hampers automatically: one product per category, no duplicates.
- Produce a branded PDF quote in minutes.
- Keep running costs low, with nothing to host and no data exposed.
The solution: EasyGift, an AI gifting platform on Amazon Bedrock
A sales person types the client’s request in plain language, or uploads a photo. EasyGift returns ranked products within budget, builds a hamper ranked to make good use of the budget, and exports a branded PDF quote.
Typed in plain language, or a photo
Budget is a hard filter
One product per category, no duplicates, ranked to use the budget well
Ready to send to the client
Key design choices
- Managed AI, nothing to host
Amazon Bedrock provides Titan Text Embeddings V2, Titan Multimodal Embeddings G1 and Amazon Nova Lite. DevOps TechLab chose Bedrock over hosting models.
- One database for products and meaning
Aurora Serverless v2 with pgvector holds both the product data and the vector search index, so there is one database to run.
- Fully serverless
Amazon API Gateway and AWS Lambda run the application, so cost follows usage.
- Built to cost little when idle
Aurora Serverless v2 scales down to its floor when idle, and VPC endpoints replaced a NAT gateway that would have cost about US$35 a month.
Security and governance
Customer data and pricing stay inside the client’s own AWS environment, with no AI keys in the code and no database exposed to the internet.
- Secrets at runtime
Credentials come from AWS Secrets Manager when needed; there are no AI keys in the code.
- Private database
The database is reachable only through Amazon RDS Proxy and an AWS Systems Manager bastion, never from the internet.
- Private network paths
VPC endpoints keep traffic to AWS services off the public internet, and S3 buckets are protected.
- In the client’s own account
Everything is deployed with AWS CloudFormation in Arham Technosoft’s own AWS account in Mumbai (ap-south-1).
Architecture on AWS
| AWS service | Role in the solution |
|---|---|
| Amazon Bedrock · Titan Text Embeddings V2 · Titan Multimodal Embeddings G1 | Understands requests and photos and turns products into searchable vectors |
| Amazon Bedrock · Amazon Nova Lite | Image understanding and AI suggestions |
| Amazon Aurora Serverless v2 · pgvector | Product data and vector search in one database |
| Amazon API Gateway · AWS Lambda | Serverless application and APIs |
| Amazon RDS Proxy · AWS Systems Manager | Private, managed database access |
| AWS Secrets Manager · VPC endpoints · Amazon S3 | Secrets, private networking and protected storage |
| AWS CloudFormation | Repeatable deployment in the client’s account |
How we delivered
- 01Discover
Assessed 854 products from 12 vendors, confirmed the need for both text and image search, and chose Amazon Bedrock over hosting models.
- 02Design
A fully serverless architecture keeping relational and vector data in one Aurora database, reviewed for cost before the build.
- 03Build
Natural-language and photo search with the budget as a hard filter, the hamper builder and branded PDF quote export.
- 04Secure & optimise
Private database access, secrets at runtime, VPC endpoints and scale-to-floor database, deployed with CloudFormation.
Business outcomes
A request that used to need manual catalogue searching and quote formatting now becomes a client-ready quote in minutes.
- Quotes in minutes
A request typed in plain language or shown as a photo becomes ranked products, a budget-compliant hamper and a branded PDF quote in minutes.
- Every quote on budget
The budget is a hard filter, so no suggestion or hamper goes over it.
- Low, predictable running cost
The full environment runs at roughly US$90–130 a month.
- Nothing to host
No AI infrastructure to manage and no AI keys in the code.
- Secure by design
No database exposed to the internet, deployed in the client’s own AWS account.
What’s next
The same platform can take on more of the catalogue and new vendors without changing the architecture. DevOps TechLab continues to advise Arham Technosoft on further AI-driven improvements to its sales process.
DevOps TechLab is an AWS Advanced Tier Services Partner with the AWS Generative AI Competency and AWS SMB Competency. We design, build and run cloud and generative AI solutions on AWS for growing businesses, with 265+ cloud environments under management.
Want AI like this working for your sales team?
Tell us where quotes, searches or proposals slow your team down. We will suggest a small first build on Amazon Bedrock, with the running cost estimated up front.
