Founder, builder & writer

Useful technology.
Expressive websites.
Words that connect.

I’m Anamika Singh, founder of Talent Vector Lab. I build AI applications and websites, bringing together practical technology, colour and storytelling.

Explore my projectsView my services

AI-assisted development. Case studies grounded in inspected code. Live assistants and synthetic demo data are clearly identified.

Explore the AI work.

Every case study explains the problem, architecture, implemented behavior and what remains to be done. The four earlier AWS projects are preserved and presented through live AI assistants over synthetic data.

Live demo · RAG + tool-use extension

RxDx knowledge assistant

The original assignment: “First has to be RAG based. Develop a QnA bot for RxDx for now.” The result retrieves clinic evidence before answering, with sources readers can inspect.

Read the case studyTry the live bot

A hybrid BM25 + embedding pipeline searches 427 chunks built from 343 public records. A separate navigator uses exact doctor filters for branch, day and language.

Launch evidence: 88 tests passed; retrieval hit@8 of 100% and MRR of 0.889 on 30 curated questions. These measure retrieval, not general answer accuracy.

FastAPIBM25 + embeddingsNova LiteVercel + OIDC

Try live · Synthetic demo data

FundBot supporter workflows

Connect supporter questions to an NGO knowledge base and build a segmented outreach workflow around donor data.

Read the case studyTry the live assistant

Existing Lambda source implements knowledge-base retrieval, conversation history, donor segments and an SES email path. The live demo retrieves a synthetic FAQ and drafts replies for fictional supporters using Nova Lite.

Verified in deployed source: chat, donor and email handlers. The knowledge base and vector resources exist. No real campaign is sent by this portfolio.

Bedrock KBLlama 3DynamoDBSES integration

Try live · Synthetic demo data

Post-discharge assistant

Organize conversation history, patient information and follow-up tasks through a conversational agent prototype.

Read the case studyTry the live assistant

Separate Lambda tools cover patient records, medications, follow-ups and appointments. The chat implementation stores history in DynamoDB and caps its tool loop.

Verified: chat plus four tool functions, three tables and cloud resources. Tool contracts need integration testing before a live care rollout.

Nova ProTool routingSession storageServerless workflows

Try live · Synthetic demo data

Clinical / FHIR tools

Retrieve a dated clinical observation from a structured bundle using its code, rather than treating the record as a block of free text.

Read the case studyTry the live assistant

The lookup code parses Patient and Observation resources, maps supported terms to LOINC codes, selects the latest result and extracts component values. The separate chat prototype has stub tools.

Verified: the FHIR Lambda parser, S3 integration and clinical tool schemas. This is prototype work, not a certified hospital-system integration.

FHIR JSONLOINC mappingsPythonLambda + S3

Try live · Synthetic demo data

AID clinician dashboard

A patient-context interface for an agent that can select relevant record and evidence tools, then organize a sourced brief.

Read the case studyTry the live assistant

The existing dashboard connects an API adapter to AgentCore. The audited tool source includes FHIR lookup, guideline retrieval, medication recall checks and external evidence queries.

Verified: dashboard HTML, Lambda adapter, AgentCore entry point and tools. The public assistant uses Nova Lite with synthetic record tools; no clinical validation is claimed.

AgentCoreNova ProEvidence toolsDashboard + API

Colour, words and a place on the web.

I enjoy shaping a website’s personality through its palette, language and flow. These two newly launched sites show how I approach different audiences.

Colour + type direction

Talent
Vector Lab

AI talent.
Human ambition.

Newly launched · My founder-led business

Talent Vector Lab

I founded Talent Vector Lab and built its website for AI recruitment, talent management, upskilling and mentorship. The design uses warm cream, dark ink and yellow accents, with direct language for teams and professionals.

The site connects service information, role descriptions and a clear enquiry path, with dedicated upskilling and mentorship pages.

Business websiteService structureBrand languageResponsive styles
Visit Talent Vector Lab

Colour + type direction

Story Lab

A little colour.
A world of stories.

Newly launched · Storytelling website

Story Lab at aikid.in

I built a playful home for children’s storytelling: birthday story parties, online sessions and age-specific experiences. Expressive typography, a colourful palette and warm language give it a different voice from a business website.

The page brings together founder introductions, programme details, parent FAQs and links for enquiries and bookings.

Creative websiteStorytellingProgramme pagesEnquiry journeys
Visit Story Lab

View my services.

Bring a website, information or workflow problem. We’ll define the audience, purpose, scope and checks that make the result useful.

RAG knowledge assistants

Question-answering over approved websites, documents and FAQs. Source preparation, chunking, retrieval, grounded responses and inspectable citations.

Typical deliverables: indexed corpus, assistant interface, evaluation questions and setup documentation.

Workflow agents and automation

Connect natural-language requests to deterministic tools, structured records and controlled actions. Design drafting and human approval as explicit workflow steps.

Typical deliverables: tool schemas, workflow prototype, validation checks and handoff behavior.

AI interfaces and cloud integration

Build usable frontends and APIs around models and data. Integrate AWS services, model adapters and hosting with scoped server-side access.

Typical deliverables: working demo, API integration, deployment configuration and operational notes.

Structured data prototypes

Turn structured JSON and domain records into queryable tools. Demonstrate context selection and clear field-level provenance using approved or synthetic data.

Typical deliverables: parsing and lookup logic, example fixtures, tests and a reviewable interface.

Website design and development

Business websites and creative landing pages with deliberate colour, typography, content structure and clear enquiry journeys.

Typical deliverables: page design, responsive implementation, contact links and deployment handoff.

Website language and storytelling

Shape the voice of a website around its audience: service explanations, programme descriptions and navigation that helps people find their next step.

Typical deliverables: page copy, content hierarchy and a consistent brand voice.

Discuss your project on LinkedIn

How we would work together.

Start with a defined task and a reviewable prototype. Expand the system when the evidence and requirements justify it.

Define the problem

Identify users, representative questions, permitted data, budget and the behavior that would make the project useful.

Build and demonstrate

Connect data, retrieval or tools and the interface. Test sample scenarios and make implementation choices visible.

Verify and hand over

Check deployed behavior, document limitations, define refresh and monitoring responsibilities, and prepare the project for its next stage.

About my work

I’m Anamika Singh, founder of Talent Vector Lab. My work brings together AI applications, website creation and storytelling.

I love playing with colour and language: mixing, blending and finding the voice that belongs to a project. I’m also writing a book, currently a work in progress.

I use AI-assisted engineering tools and inspect the resulting behaviour, trade-offs and evidence. The projects here show both my creative choices and the implementation behind them.

View my LinkedIn profile

What this portfolio makes clear

  • Working behavior and prototype scopeRxDx has a live bot. All four additional projects have live Nova Lite chat with synthetic data, separate from the audited older backends.
  • Evidence over invented resultsNo fabricated client history, certifications, revenue gains or clinical validation.
  • Practical delivery choicesSource traceability, bounded tools, server-side credentials and cost-aware demonstrations.

A case study you can share.

The detailed RxDx deck covers the original RAG assignment, architecture, build choices, AWS/Vercel deployment, testing and lessons. LinkedIn-ready copy is available below.

Download the 31-slide RxDx deckOpen LinkedIn-ready copy
Copy a short Services description

Professional project enquiries.

For clients, founders and teams with a genuine project: share your professional profile, the task, scope and delivery expectations. Unsolicited promotions, unrelated requests and spam will not receive a response.

Message me on LinkedInPrepare a project briefTry the RxDx bot