Amit Gupta

SU MM A RY

Experienced AI and data science

specialist with 5+ years in developing
scalable solutions in telecom, DATA SCIENTIST|AI/ML SPECIALIST
banking, insurance and R&D sectors.
Have delivered projects following
I am passionate about building AI solutions to solve defense and
Agile methodology, ranging from
security problems. With an academic background in
recommendation systems, time-series
Mechatronics engineering, I gained in-depth experience in AI,
forecasting, NLP text classification
Data Science, Computer vision, modeling and simulation through
and Generative AI using Python,
my studentarbeit, masterarbeit and professional work experience.
Pytorch, SQl, LangChain and Flask.
This combined with my passion drives me to contribute to the
Proactiveness and self-driven
cutting-edge solutions in current highly dynamic landscape.
approach keep me updated in this
highly volatile tech world.

E XP E R IE N CE
CO NT AC T

A I/ ML SP EC I AL I ST AC C E N T U RE S O LU TI O N S P VT . L TD .
• Linkedin
Nov 2024 - Present
• +91-9060211542
• Led the redesign of the company's Chatbot enhancing
• Pune, India
the AI capabilities using RAG, Semantic Kernel and
ChatGPT 4o. This resulting in 85% improvement in True
Positives.
ED UC A TI O N • Designed and developed an evaluation script leveraging
Confusion Matrix for the answers generated by the
UNIVERSITY OF SIEGEN Chatbot.
M.Sc-Mechatronics • Conducted extensive descriptive data analysis of the
2023 performance of the chatbot. Collaborated with
development and leadership teams, documented and
KIIT UNIVERSITY, Bhubaneshwar presented the results.
B.Tech-Mechanical Engg. • Deployed the solution using Python, Flask, Git CI/CD on
2017 Azure Cloud.

PROJECTS AND THESIS

SE N IO R NL P E NG I N EE R T U R ABI T S O L U TI O NS P V T. L T D.
• Studentarbeit: Explainable May 2024 – Oct 2024
Pneumonia detection using
CNN. • Developed greetings detector leveraging Machine
• Masterarbeit: 3D human body Learning models such as XGBoost, Bert with 94%
reconstruction using accuracy. Also enhanced the chatbot performance by
optimization function. 30% improvement in precision.
• To achieve above milestone, trained and finetuned NLP
models such as XGBoost with TFIDF, BERT, KeyBERT and
Llama 2 7B. Used PEFT and LoRA methods for finetuning.
• Conducted validation on the model performance using
A/B testing and Hypothesis testing(Z-test).
• CI/CD using Git on Azure Cloud.
SK IL L S
P RO DU CT E N GI N E ER V- L AB S P VT . L TD .
Feb 2023- April 2024
• Python, Pandas, Numpy, SpaCy,
PyTorch, Sci-Kit learn, Matplotlib, • Developed and deployed the state-of-the-art solution to
extract .eml file containing excel sheets with customer
Seaborn, Plotly
information for an insurance client, we achieved 92%
• Flask, Semantic Kernel, accuracy compared previous solution with 27%. Also
LangChain, LangGraph implemented LLM RAG based solution for the same.
• The solution contained 4 different layers of text
• Azure, AWS ML
classifications leveraging Machine learning, text analytics
• SQL, PostgreSQL, MongoDB, Milvus using SpaCy, OCR.
(Vector DB)
• Git, CI/CD, Jupyter, PyCharm, VS
code RE S E AR C H E NG I N EE R - D AT A S CI E NT I ST ( W H I L E DO I NG
M. SC )
Ciena, India, Mar 2022- Oct 2022
CERTIFICATIONS
• Generative AI – • Developed and trained forecasting models for predicting
telecom network outage. Also worked on the data
DeepLearning.Ai
analysis for building a recommendation system on
• Data Science telecom data.
• AWS ML – Coursera • Brainstormed with the team using techniques such as
RCA, anomaly detection using 95% CI, PCA and clustering
• MLOps – Udemy
such as KMeans and DBSCAN.
ENSUN GmbH, Germany, Oct 2022 – Jan 2023
REFERENCES – Letters are available • Trained and deployed a multi-class text classification for
B2B tech scouting startup.
upon request.
• Built an ETL pipeline. Scrapped data from internet,
• Benjamin Hill – CEO, ENSUN cleaned data using regex, SpaCy and loaded to on
GmbH premise DB storage in MongoDB based on the
classification done.
• Hasan Abu-Rasheed –
Research Assistant, University
of Siegen.
PROGRAMMER ANALYST COGNIZANT TECHNOLOGY SOLUTIONS
PVT. LTD.
Nov 2017- Aug 2019

• Built a PoC, Product recommendation system for a US

based retail client.
• Mostly based on comparative data analysis of customer
ordering history a recommendation model was built. We
used time of ordering and product ordered.