2018
Researcher
Yale University
Research at the intersection of neuroscience and computation, exploring the foundations of intelligence.
- Yale Inference Project
- Causality · Entanglement
Hi, I'm Gaurav
Ivy League labs taught me to ask questions; AI taught me to find answers. I build data & ML systems and write about science, technology, and the human side of progress.
My Journey
Four years of shipping AI to production, from healthcare to hospitality to enterprise analytics. Swipe through the chapters.
-
Research -
Education 2021
Bachelors in Computer Science
University of Delhi
Foundations in algorithms, data structures and software engineering.
- Minor in Mathematics
- 8.32 / 10
-
Education 2022
Masters in AI & Machine Learning
UC Berkeley × Plaksha University
Tech Leaders Program: advanced machine learning, deep learning and AI ethics.
- Outstanding Capstone Award
- Cum laude · 3.54 / 4
-
Engineering Apr — Jun 2022
Data Scientist
HRS Group · Mohali
Scalable, real-time dashboards over AWS Orbit Workbench, Athena and data lakes.
- 95% less manual reporting
-
Engineering Aug 2022 — Jun 2024
Associate Data Scientist
Minfy · Gurugram
Led a healthcare AI platform for discharge reporting and chronic disease management.
- Reports: 30 min → under 2 min
- 40% shorter consultations
Healthcare AI & chronic disease management
Problem
Healthcare teams needed leaner operations and a way to put medical data to work for chronic disease care.
What I built
- An AI platform generating HL7 FHIR-compliant discharge reports with an optimised Mistral 7B on AWS.
- Structured extraction and anonymisation with Amazon Textract and Comprehend.
- A Patient 360 dashboard and disease progression tracker on Amazon HealthLake, Flask, Athena, QuickSight and Angular, HIPAA compliant.
Impact
Discharge reports went from 30 minutes to under 2, and consultations got 40% shorter, freeing time for preventive care.
Mistral 7B · HL7 FHIR · HealthLake · Textract · HIPAA
-
Engineering Jun 2024 — Aug 2025
Data & ML Engineer
HRS Group · Mohali
Built a hotel procurement copilot that turned multi-day research into a conversation.
- 80% faster analysis
- 60% less consultant reliance
Hotel Procurement Copilot
Problem
Travel managers spent days researching hotel procurement and leaned on costly external consultants.
What I built
- LLMs on AWS Bedrock with a ReAct agent and retrieval-augmented generation, querying and analysing data in real time.
- A serverless pipeline on AWS Lambda, Amazon ECR and API Gateway for on-demand, scalable compute.
- A chat API delivering procurement insights, risk assessments and data-driven hotel recommendations.
Impact
Analysis time cut by 80%, from days to hours, and consultant reliance down 60%, lowering operating costs and enabling faster, better-informed sourcing decisions.
AWS Bedrock · ReAct · RAG · Lambda · API Gateway
-
Engineering · Now Sep 2025 — Present
AI Engineer
MicroStrategy · Pune
Building the LLM platform layer and prescriptive analytics for Fortune 500 companies on MicroStrategy ONE.
- Prescriptive analytics, built from scratch
- Enterprise prompt platform
- 5 enterprise personas
AutoPrompt — enterprise prompt management
Problem
Prompt templates were scattered across codebases, configs and notebooks, with no traceability, constant version conflicts and unsafe production deploys.
What I built
- A centralised, self-hosted platform on FastAPI, PostgreSQL and AWS for enterprise-scale LLM operations.
- Immutable prompt versioning over a multi-tenant hierarchy (Customer → Provider → Model → Agent), with governed promotion, approval gates and audit trails.
- RBAC, JWT and API-key authentication, and AWS SES reviewer notifications.
- A production Python SDK (sync and async) with caching, automatic token refresh and offline fallback, plus GitHub Copilot-assisted prompt authoring with SSE streaming.
Impact
One governed home for every prompt and lower production regression risk. Self-service authoring and a low-friction SDK drove adoption and faster iteration across AI teams.
AutoThink — prescriptive analytics for AutoBot 2.0
Problem
AutoBot 2.0 covered descriptive, diagnostic and predictive analytics, but stopped at insight. Users still had to work out what to do next.
What I built
- An end-to-end prescriptive analytics layer on MicroStrategy ONE, architected from scratch.
- A multi-stage agentic LLM pipeline: eligibility classification, persona detection, dual-path query decomposition, market intelligence synthesis and persona-aware recommendations.
- Adaptive persona intelligence for five roles (Executive, Data Analyst, Financial Analyst, Business Manager, Operations Manager) that tunes vocabulary, metrics and time horizon.
- Dual-path execution: structured sub-queries through MicroStrategy’s semantic SQL layer, unstructured ones through vector search across enterprise documents.
- A real-time market intelligence engine for industry-aware, geo-temporal context.
Impact
True next-best-action recommendations inside an enterprise BI workflow. Rival platforms such as Databricks and Snowflake focus on insights and conversational interfaces, so this closed a real capability gap and set the product apart.
FastAPI · PostgreSQL · AWS · Agentic LLMs · Vector search
Credentials
Certified across AWS and Google Cloud.
- Machine Learning Specialty Amazon Web Services
- Big Data Specialty / Data Analytics Amazon Web Services
- Solutions Architect Associate Amazon Web Services
- Associate Cloud Engineer Google Cloud
- Cloud Digital Leader Google Cloud
Toolkit
- AI / ML
- Generative AI, Agentic AI, LLMOps, Recommendation Systems, A/B Testing
- Languages & tech
- Python, SQL, Docker, Kafka, Apache Spark, Heroku, Gradio, Flask, FastAPI, Git, Bash
- Databases & search
- Amazon DynamoDB, Amazon OpenSearch, Amazon Athena, PostgreSQL, MongoDB
- MLOps & data
- ETL pipelines, feature engineering, SageMaker and Lambda deployment, AWS Glue, SQS / SNS / SES, vector databases, CloudWatch monitoring
- Tools & BI
- MicroStrategy, GitLab CI, GitHub Actions, Amazon QuickSight
Let's Build Something Together
Always open to discussing new ideas, collaborations, or just chatting about the future of AI.
Get in Touch