graph by datavruti
mygraph.id/abhi12ravi

Abhiram Ravikumar

Senior Data Scientist

Bengaluru, IndiaBuilding since Jul 2016 (10+ yrs)
Headline

From speeding up SAP's enterprise apps to scaling a Medicare RAG chatbot to 50M+ members, my work consistently compounds: faster inference, lower cost, and higher accuracy at growing scale.

6direct reports managed

direct reports

3.7baseline NPS score

out of 5

4-person teamteam size
10client adoption

clients (MarTech & FinTech)

Represented Roles

Specialized Functions & Roles.

Engineering capabilities and specialized roles demonstrated across real production work.

Senior Data Scientist

Primary Role

Applied AI / ML Engineer

NLP / LLM Engineer

MLOps Engineer

Signals

Live in Production.

Each signal ties back to a specific, evidenced moment.

Built

5 moments

Systems shipped end-to-end, from schema to deploy.

  • •Built web search integration using Bedrock Mantle
  • •Shipped Cost Agent in Avery using Python, LLMs, and RAG
  • •Built Bodhi SLM Factory for enterprise fine-tuning, adopted by 10 clients across MarTech and FinTech, cutting training time 2x and deployment time 3x

Solved

4 moments

Gaps closed - false positives, edge cases, dead ends.

  • •Cut healthcare testing effort 3.4x with an AI-driven framework that automated test data creation and test runs on Kafka pipelines
  • •Boosted labeling precision from 75% to 92% on 270M FMCG posts while cutting costs 90% vs. human labeling, using Mistral on AWS Bedrock for its stronger performance on European languages, validated against a representative human QA sample
  • •Cut transformer inference times 200% at Collinson, helping lift NPS from 3.7 to 4.26 out of 5

Scaled

2 moments

Taken from prototype to real production volume.

  • •Scaled a Medicare RAG chatbot to 50M+ members with 500K+ monthly active users and >92% intent accuracy
  • •Scaled a merchant recommender system to process 5M+ daily transactions, cutting latency from 1.5s to 450ms

Learned

2 moments

New tools and depth picked up on the job.

  • •Developed people manager skills while managing 6 direct reports, through reading and the People Manager essentials course
  • •Pioneered synthetic data pipelines for scene text recognition
Growth trajectory

Abhiram's career, mapped.

A timeline of milestones captured directly from real work.

  1. Began building enterprise software

  2. Pivoted into applied ML research

  3. Formalized skills in data science

  4. Scaled NLP into production systems

  5. Took ownership of LLM initiatives

  6. Stepped into enterprise-scale AI leadership

Stack

Daily tools.

What Abhiram actually reaches for, pulled from the tags attached to captured moments.

Work

Key Work By Role

What they built, shipped and owned at each role, with impact numbers where there are any.

Publicis Sapient

Senior Data Scientist · 2025 - Present
Apr 2025 - Oct 2026

Avery Conversational Companion (UHG Optum)

Enterprise RAG-based chatbot serving 50M+ Medicare & Medicaid members with high-precision intent classification and GenAI workflow productionization.

>92% intent accuracy in production500K+ monthly active users~40% reduction in deployment cycle time
Stack:RAGFastAPIGPTKubernetesGitHub ActionsBedrock
View case study
Apr 2025

AI Test Stream (US Healthcare)

AI-driven, automated testing framework on Azure AKS and ML Workspace for a Fortune 5 Healthcare client.

3.4x reduction in manual testing effort4.2x faster testing cycles
Stack:Azure AKSML WorkspaceTest AutomationAI
View case study
Apr 2025

Bodhi SLM Factory

Platform for fine-tuning and deploying small language models on proprietary enterprise data.

2x reduction in model training time3x reduction in client deployment timeAdopted by 10 clients across MarTech and FinTech domains.
Stack:SLMFine-tuningMLOpsNLP
View case study

Ai Palette (Reflion Tech)

Senior Data Scientist · 2024 - 2025
May 2024

Contextual Tagging project

Contextual tagging system quantizing and fine-tuning Phi-3-Instruct and DeepSeek-R1 for ingredient, product, and theme tagging from social media posts.

2.7x reduction in inference cost via CPU optimizationTrained on 50K instruction pairs
Stack:Phi-3DeepSeek-R1QuantizationKnowledge Graphs
View case study
May 2024

Historical Cleanup project

LLM-based analysis of 270M FMCG social media posts using Mistral, replacing third-party human labeling with foundational models.

Precision improved from 75% to 92%90% cost savings vs. human labelingCollaborated with a 4-person team, including a DS intern, a lead data scientist, and 2 data engineers who automated DAG pipelines in Airflow.
Stack:MistralAWS BedrockAzure MLLLM
View case study

Collinson

Cloud Machine Learning Engineer · 2022 - 2024
Apr 2022

Text Analysis (TEXA) project

NLP and topic modeling pipeline analyzing 500K+ airport lounge customer reviews using BERTopic and GPT models.

200% reduction in inference timeBoosted NPS by 15%Analyzed 500K+ customer reviewsStarted from a baseline NPS of 3.7 out of 5 before achieving a 15% boost.
Stack:HuggingFaceAWS SagemakerBERTopicGPT
View case study
Apr 2022

Merchant Recommender (MER) project

Recommender system pipeline processing 5M+ daily transactions with automated ETL, model orchestration and brand extraction.

Processed 5M+ transactions dailyLatency cut from 1.5s to 450msIdentified 250+ brands/merchants
Stack:KubernetesTerraformElastiCacheAWS ComprehendHDBSCAN
View case study

King's College London

Research Software Engineer · 2021 - 2022

Deployed an NHS decision support system on AWS

Deployed an expanded decision support system on AWS for the NHS, automating processes to reduce operational overhead.

40% reduction in deployment time25% savings in operational time
Stack:AWSDecision Support SystemsAutomation
View case study

SAP Labs

Machine Learning Research Fellow · 2018 - 2018

Pioneered synthetic data pipelines for scene text recognition

Pioneered synthetic data creation and annotation for a scene text recognition inference service using Synth-Text, and designed, developed and deployed TensorFlow-based CNN models on AWS.

Stack:TensorFlowCNNSynthetic DataAWS
View case study

Sped up SAP's Subcontracting Cockpit app by 90%

Resolved performance bottlenecks in the Subcontracting Cockpit app within the SAP Core Data Services (CDS) stack, and developed B2B web applications on SAP S/4HANA using SAP UI5, Fiori and JavaScript.

Reduced load time from 30 seconds to 3 seconds (90% faster)
Stack:SAP CDSSAP UI5FioriJavaScript
View case study

SAP Labs

Software Developer · 2016 - 2020
Outside Work

Side projects, open source & publications.

Independent builds, open-source contributions, technical writing, and experiments beyond the day job.

Experiment

People Management Skills

Working on people manager skills by reading books on management and taking the People Manager essentials course.

Present
Learning

Formal record.

  • MSc Data Science
    King's College London
    2021 - 2022
  • BE (hons) Information Science
    PES Institute of Technology
    2012 - 2016
Activity heatmap

A record of momentum.

A combined view of Abhiram's activity across the tools used daily.

LessMore

Profile last updated today.

Liked Abhiram's graph?

Build yours today.

Start with your resume