Latest
  • Zero to one, and one to N
  • Responsible AI at scale
  • Agentic procurement workflows
  • Cloud infrastructure on GCP and AWS
  • Certified AI Builder
  • Published at ICRAS 2024
  • Open to interesting problems
Vol. I · No. 290 Hyderabad edition

Human in the Loop

AI that’s fit to ship

Front page · Responsible AI

Ankith builds responsible AI, from the ground up

A senior software engineer at a Hyderabad startup, leading backend and cloud engineering for an AI-native procurement platform, with a focus on scalable systems, production engineering and responsible AI.

By Sai Ankith PotapragadaSenior Software Engineer, HolmesAIHyderabad, India

I’m a founding engineer who builds AI products from the ground up, taking them from zero to one and then from one to N, with a focus on cloud infrastructure, scalable systems, production engineering and responsible AI. I like the work that sits between a promising model and a product an enterprise can depend on.

At HolmesAI I lead backend and cloud engineering for an AI-native procurement platform, and at a startup that means owning a lot of the stack, from cloud infrastructure to the services on top. That covers the core AI infrastructure and backend services behind agentic procurement workflows, built to support enterprise-scale operations.

My background spans research on robotics and conversational AI at a research lab, where I co-authored a paper on child-robot interaction, data science work on retrieval-augmented generation, and backend and cloud engineering for enterprise AI.

Continued in Selected Work →

Selected Work

Three stories

Platform · Enterprise procurement

An AI-native procurement platform, built from the ground up

I lead backend and cloud engineering for this multi-tenant procurement platform at HolmesAI. I designed its cloud infrastructure from scratch, set up secure CI/CD pipelines and built the distributed backend that supports enterprise deployments.

On top of that sit agentic workflows for document analysis, supplier intelligence and risk assessment. They take a supplier from contract ingestion and onboarding to a risk view across legal, financial, compliance and operational dimensions, so procurement teams spend less time on repetitive checks.

Agentic AI · FastAPI · GCP · AWS

Project · Healthcare AI

DOCter: healthcare answers grounded in your own reports

A multi-turn chatbot that gives personalised healthcare insights by processing the medical reports a user uploads. It pairs language models with grounded document processing so answers stay relevant to the report in front of it.

RAG · LLMs

Research · Published

A conversational robot that teaches hand hygiene

HaKsh-E, a conversational social robot, taught hand hygiene to 36 children aged 7 to 11. They trusted it alone and in groups: groups drew more interaction, solo sessions larger learning gains.

Conversational AI · Social robotics · ICRAS 2024

The Stack

What I build with
AreaInstruments
AIAgentic systems and RAG: multi-step orchestration, tool calling and evaluations across multiple providers, guardrails, AI ethics and data privacy
BackendPython, FastAPI, distributed and scalable systems
CloudGCP, AWS, on-prem, secure CI/CD

Accomplishments

Awards, a paper and credentials
  • Paper

    Exploring Child-Robot Interaction in Individual and Group Settings in India

    Presented at the 8th International Conference on Robotics and Automation Sciences (ICRAS 2024) in Tokyo.

  • Credential

    Building AI Agents and Agentic Workflows Specialization

    IBM certified.

  • Rank

    Campus Rank Awardee

    Amrita School of Computing. Master of Computer Applications, graded 9.45 out of 10.

  • Scholarship

    Merit Scholarship

    Awarded by Amrita University to the top quarter of eligible students in the Master’s program.

  • Leadership

    Chairperson, IEEE Education Society

    Chaired the society at Amrita Vishwa Vidyapeetham, Amritapuri.

Certificate Awarding Ceremony, Tokyo, Japan
Certificate Awarding Ceremony, Tokyo, Japan.

Letters to the Builder

Cache You Around

Building with AI, agents or the infrastructure underneath them, or simply curious where it all goes? Bring an idea, a question, or a half-formed thought. My inbox is open.