Akarsh Raj

AI Researcher

Erlangen, Bavaria, Germany6 yrs experience

AI ML PractitionerHighly Stable

Key Highlights

7+ years of experience in Data Science.
Expert badge holder in Kaggle, ranked top 1%.
Led multiple high-impact AI projects in healthcare and finance.

Stackforce AI infers this person is a Data Science expert with a focus on Fintech and Healthcare solutions.

Contact

Skills

Core Skills

Agentic AiGenerative AiNlpOptimizationFraud DetectionChurn Management

Other Skills

Retrieval-Augmented Generation (RAG)LangGraphLangChainPineconeTeam BuildingGPTText AnalyticsAutoencodersIsolation ForestLinear ProgrammingMachine LearningAPI DevelopmentSentiment AnalysisNatural Language ProcessingPython

About

Currently pursuing Masters in Artificial Intelligence (until April'26). Available for Full time (FTE) roles from 1st May. Akarsh has more than 7 years of experience as a Data Scientist. He last worked with Digital Intelligence & Robotics team at Evalueserve and led an internal Products team. He developed and led GPT (NLP) based products , which aimed to increase the efficiency of employees within Evalueserve using machine learning and software engineering. Previously he worked with Advanced Data Science(ADS) team at ZS Associates. Prior to that, he has worked with Xceedance as Data Scientist. He has worked and delivered several projects across different industry verticals. He was previously involved with leading pharmaceutical companies. In past, he collaborated with the biggest giants in tech and insurance industry and helped them with their Data science components. Apart from his day job, he takes huge interest in community meetups and believes in giving back to the community from where he himself learnt a lot and still learning everyday from it. Also, Akarsh holds the Expert badge in Competition tier at Kaggle and is ranked among top 1%. Akarsh's interests include financial markets, cryptography, solution design of real-world data science problems. Apart from technology, he has great interest in Formula One Racing, Soccer, Cricket and Tennis and loves to play these games too. Two cents from Akarsh Raj "Treat your model like your new born child. Make it learn the good things, things which make sense and teach it to be open and broad minded."

Experience

6 yrs

Total Experience

2 yrs

Average Tenure

Current Experience

Siemens healthineers

Agentic AI Engineer

Jul 2024 – Present · 1 yr 10 mos · Forchheim, Bavaria, Germany · Hybrid

FinCom AI :-
>> Architected a financial Q&A RAG system using a multi-agent architecture built with LangGraph and LangChain, leveraging Pinecone for high-dimensional vector retrieval.
>> Engineered a deterministic execution layer that generates Python code to prevent hallucinations, integrating Chain-of-Thought (CoT) reasoning and model traces to ensure accuracy and transparency for industrial workflows.

Agentic AIRetrieval-Augmented Generation (RAG)

Evalueserve

Data Science Lead - DIR

Apr 2022 – Jul 2023 · 1 yr 3 mos · Gurugram, Haryana, India

>> Infinity AI : Led, designed & developed a RAG application tool on top of GPT models for abstract Q&A, information extraction, and more from personal or organizational documents using natural language. Improved firm's research capabilities by over 30% (w.r.t TAT and Quality). Also implemented redaction of PII data, processing of tabular data and scanned PDFs using table localization & OCR techniques
>> Canvass AI – Led, designed & developed a NLP system for researchers which web crawled google results (SERP-API) based on a query and generates various text analytics (NER, n-gram analysis, sentence level clustering, machine comprehension, etc.) on the results,
increasing researcher's efficiency by over 40%
>> Led a team of 10+ Data Scientists & Software Engineers; Reported to CXO leadership to develop quarterly roadmaps based on impact,
effort, and efficiency.

Generative AITeam Building

Zs

2 roles

Data Science Associate Consultant - Advanced Data Science Group

Promoted

Dec 2020 – Apr 2022 · 1 yr 4 mos

> Copay Anomaly Detection: Built an anomaly detection engine using autoencoders and isolation forest to identify fraudulent pharmacies and prescribers, uncovering $5M+ in annual fraudulent transactions; enhanced model interpretability with SHAP explanations
> Clinical Trial Optimization: Designed and implemented a linear programming solution to optimize study duration via clinical trial site selection while satisfying thousands of constraints
> Taxonomy Modeling: Designed and implemented an AI-driven knowledge graph on top of notes from medical field agents data using BERTopic. Built a hierarchical taxonomy and deployed the structured ontology in Neo4j to enable semantic querying, relationship mapping, and intelligent data insights.
> Time Series Forecasting: Forecasted new-to-brand (NBR x) drug numbers using Prophet, integrating COVID-19 impact into the model for improved accuracy
> Handled multiple client accounts, ensuring alignment with Fortune 10 and S&P 500 requirements.
> Fashioned presentation decks and tabulated documentation for presentation to CXO leadership & Partners, on a regular basis

OptimizationNLP

Data Science Associate - Advanced Data Science Group

Oct 2018 – Dec 2020 · 2 yrs 2 mos

> Clinical Trial Optimization & Predictive Analytics: Architected a clinical trial optimization
engine using Linear Programming to satisfy thousands of operational constraints and
spearheaded Time Series Forecasting (Prophet) for NBRx metrics, successfully
integrating COVID-19 impact variables.
> Receipt Detection using Deep Learning: Developed an object detection model using Faster-RCNN to detect receipts in the image; Implemented Transfer-Learning on pretrained COCO weights and achieved a MAP of 85 % in production

Xceedance

2 roles

Data Scientist

Jun 2017 – Sep 2018 · 1 yr 3 mos · Gurgaon, India

> Fraud Detection Model - Coded a model to predict whether there will be a claim in the first term of insurance itself - for one of the largest US based insurance firm which reaped them multi-million profit, annually. Also developed an API for the same using Flask
> Coded a product valuation tool using various ML algorithms, while also focusing on machine learning interpretability(MLI), XAI and FAT ML. It outperformed all the existing valuation tools in client's system
> Developed a customer-lapse model (Churn) for an Australian based insurance firm which analyzed customer's sentiment and all other information to predict their lapse
> Introduced Text Field Classification System assisting underwriting process, minimizing process time from 20 minutes/entry to an instant using OCR, sequence models and various NLP techniques
> Worked extensively on OCR to extract information from readable and non-readable(scanned) PDFs
> Handled all the weekly and daily client calls and received numerous appreciation for understanding client's every requirement and for exemplary presentation as well as client handling skills

Churn ManagementFraud Detection