Yash Deore

I design and ship production AI systems on AWS — for teams in healthcare, life sciences and aviation.

Two years into client work, my focus is narrow and practical: take a regulated or data-heavy problem, architect it on Bedrock, Neptune and serverless AWS, write the Statement of Work, build it with CDK, and hand it over documented and owned by the client. Nine AWS certifications, including Solutions Architect Professional and the Generative AI Developer Professional early-adopter track, back that up.

9AWS certifications
4client platforms delivered or in flight
3industries — health, pharma, aviation
2024research accepted at QCRYPT

Client work

AlphaGutScan — sample-to-outcome microbiome pipeline

AlphaBiomics — healthcare AI, via Cloud202

A universal six-stage pipeline that turns a lab sample into a clinical outcome report, with five quality gates in between. Deployed in Frankfurt alongside the client's bioinformatics workloads on AWS HealthOmics.

  • Owned the knowledge-graph workstream on a 12-week proof of concept
  • Wrote the complete prompt library (P0–P6), from metadata intake through chatbot guardrails
  • Designed a domain-pack / frozen-skeleton architecture so the same pipeline serves new test types without re-engineering
  • Consolidated model usage onto a single Claude model on Bedrock for cost and consistency
StackAmazon Bedrock · Neptune · S3 Vectors · Bedrock Knowledge Bases · Lambda · HealthOmics · eu-central-1 RoleKnowledge-graph lead, prompt architecture

Orion — aviation intelligence platform

FlyTribe — aviation AI, via Cloud202

A GraphRAG system for aviation knowledge, scoped and documented from the ground up. The Statement of Work was written to AWS partner-funding standards, framing every deliverable as design, build and deploy.

  • Full SoW and architecture for a Neptune + vector + Bedrock Knowledge Base retrieval pipeline
  • Established the GraphRAG pattern that was then reused on AlphaBiomics — one architecture, two industries
StackNeptune · S3 Vectors · Bedrock Knowledge Bases · Lambda · CDK RoleSolution architect, SoW author

Life sciences AI workflows and knowledge graph

Qubitz — pharma and clinical research

Three productised workflows built to be licensed many times over rather than rebuilt per customer, now extending into a shared biomedical knowledge graph that all three can query from a single retrieval layer.

  • Clinical-trial patient recruitment — TrialGPT-style matching on the Chia and TREC 2021 datasets, with a metabolic-disease demo
  • Pharmacovigilance — adverse-event analysis over FAERS with MedDRA coding
  • Regulatory submissions — NDA dossier assembly in CTD structure, tested against a synthetic drug dossier and lab reports
  • A GraphRAG agent deployed on Bedrock AgentCore, giving each workflow traversal across entity relationships instead of flat vector similarity alone
  • In flight: ingestion of public biomedical catalogues into a Neptune property graph, normalising heterogeneous source schemas into one edge model so a new source is a mapping rather than a migration
  • IAM design for the Amplify and AgentCore deployment, including resolving managed-policy quota limits
StackAmazon Bedrock · AgentCore · Neptune · Amplify · Lambda · IAM RoleAI engineer, productisation

BriefBrain — AI brief-writing platform

PlatformAlt5 — brand and creative strategy, via Cloud202

A production web application that turns a guided questionnaire into a finished creative brief. I own the generative layer beneath the product — the retrieval, the grounding, the guardrails and the document pipeline that together decide whether the output is something a brand team will actually put their name to.

  • Retrieval-augmented generation over a curated knowledge base blended with live external sources, so claims in a brief resolve to a citation rather than to model recall
  • Industry-trend aggregation with clickable sources, and competitor activity summarised per sector at request time
  • Tone and audience modelling — twelve brand archetypes and generated customer avatars feed the prompt layer, so output reads as the brand rather than as the model's default voice
  • Guardrails against hallucinated, biased and copyrighted output; client content stays private to the account and is never used to train foundation models
  • Formatted PDF generation and delivery, plus ongoing production ownership — including a full rewrite of the generation path after tracing a corrupted brief back to an aggregation leak in the knowledge base
StackAmazon Bedrock (Nova, Claude) · Bedrock Knowledge Bases · Guardrails · Lambda (eu-west-2) · SerpAPI · VPC & PrivateLink · ReportLab · S3 RoleAI engineer — generative layer, production ownership

A pattern I keep returning to: build the retrieval and graph layer once, make the domain a configuration pack, and license the result to the next client. It is why the AlphaBiomics, Orion and Qubitz work share the same bones.

Experience

Jul 2024 — present

AI Engineer

Cloud202 · AWS Advanced Tier Services Partner · UK, remote

Cloud architecture and Generative AI delivery for healthcare, life-sciences and aviation clients — Bedrock, SageMaker, RAG and GraphRAG pipelines, vector databases, workflow automation and infrastructure as code. Write the Statements of Work, then build what they describe.

Aug 2023 — May 2027

Diploma in Computer Engineering

MSBTE · in progress alongside client work

What I do

Generative AI architecture

RAG and GraphRAG on Bedrock with Neptune and vector stores, prompt libraries with quality gates, guardrails, evaluation and model consolidation.

AWS platform engineering

Account foundations, least-privilege IAM, CDK v2 in TypeScript, serverless and container workloads, backups, cost controls and security baselines ready for VAPT.

Scoping and delivery

Statements of Work with sprint plans, milestones and acceptance criteria; change-request discipline; client communication from kickoff to handover.

Regulated domains

Clinical trials, pharmacovigilance, regulatory dossiers, microbiome diagnostics — data handling and documentation to match.

CI/CD and handover

CodePipeline, GitHub Actions, CDK Pipelines, runbooks and playbooks written so the client's own team can operate without me.

Certifications

May 2026AWS Certified Developer — Associate
Jan 2026AWS Certified Generative AI Developer — Professional (Early Adopter)
May 2025AWS Certified Data Engineer — Associate
Apr 2025AWS Certified Machine Learning — Associate
Mar 2025AWS Certified Solutions Architect — Professional
Jan 2025AWS Certified AI Practitioner (Early Adopter)
Sep 2024AWS Partner: Generative AI Sales
Apr 2024AWS Certified Solutions Architect — Associate
Apr 2024Data Science for Engineers
Jan 2024AWS Certified Cloud Practitioner
Dec 2023Microsoft Certified: Azure Fundamentals
Sep 2023Python for Data Science — NPTEL

Research & talks

Apr 2026

Architecting Gen AI solutions for video intelligence on AWS

Industry expert lecture, Rizvi College of Engineering.

Oct 2024

InnovateX — AI innovation and entrepreneurship

1st runner-up. A FinTech AI solution built solo on AWS with Llama models, competing against full teams.

Jul 2024

Quantum cryptanalysis of FIPS 140-3 compliant encryption standards

Research paper accepted for presentation at QCRYPT 2024.

Contact