Sakshi Tantak profile photo 1 on Tal

Sakshi Tantak

Senior Data Scientist, Acko | Deep Learning | topmate.io/sakshitantak

Bengaluru

5+ years of work experience

Acko
Mar 20242 yrs 2 mosPresent

Bengaluru, Karnataka, India

Built platform for identifying document fraud at the time of claim Built central platform that enables building any document intelligence use cases with the help of a mere configuration file. Conducting continued pre training and fine-tuning of SLMs for domain adaptation in insurance like analysing policies, and enabling JSON data interpretation using LLMs.
Vakilsearch
Jun 202310 mosMar 2024

India

Conducted continued pre-training and fine-tuning of language models ranging from older to relatively newer architectures to serve and enhance legal domain. Built a chatbot to help reduce escalations in customer experience section with LLMs using RAG
i3systems.ai
May 20221 yr 2 mosJun 2023

Mumbai, Maharashtra, India

- Worked on developing models and strategising data for Financial Underwriting platform - an intelligent platform to extract and validate data from financial documents like Salary Slips & ITR. Trained CRF-BiLSTM based architectures using flairnlp & CRF++ for NER with a very generic tagging strategy for documents where data exists in key-value pairs - Fine-tuned Clinical LongFormer for extraction of clinical & non-clinical data from discharge summaries - Fine-tuned LiLT for extraction of major sections from discharge summaries - Conducted trial to predict quality of document images using LayoutLMv3 - Fine-tuned LayoutLM, LayoutLMv2 & LiLT for extraction of PII from KYC documents - Aadhar, PAN Card, Passport, Voter ID & Driving Licence - Conducted trial for image classification by fine-tuning Vision Transformer based architectures - Worked on correcting angles of document images to prevent issues arising from disoriented images in inference of multimodal models - Built quick demos and internal annotation tools using Streamlit - Customised existing streamlit component for NER annotation/highlighting to aid highlighting entire token instead of substrings selected upon dragging - Customised existing streamlit component to convert BIO to BIOES style of annotation for NER - Working on developing a system to identify potentially fraudulent claims by using a vast array of data points extracted from documents attached with claims, using objective triggers & embeddings based search & retrieval of claims with similar to exact section occurrence in documents
i3systems.ai
Aug 202110 mosMay 2022

Mumbai, Maharashtra, India

- Trained NER models to extract Metadata from Bills, Reports & Discharge Summaries using flairnlp, SpaCy & CRF++ - Integrated Microsoft Azure’s Language Services in an existing product for medical codes like ICD10, UMLS, etc using textanalytics healthcare API - Trained NER models for extraction of PII information from KYC documents - Aadhar, PAN Card, Passport, Voter ID, Driving Licence & Insurance IDs - Experimented with a lot of models for multiple tasks
Indian Institute of Technology, Bombay
May 20215 mosSep 2021

India

Omdena
May 20213 mosJul 2021

India

NLP, Web Scraping

Sakshi’s AI Stack

hugging face
hugging face
langchain
langchain
langgraph
langgraph
openai
openai

Education

Pune Institute of Computer Technology

Bachelor of Engineering, Information Technology

Pune Institute of Computer Technology

20174 yrs2021

Join Sakshi and 10k+ other professionals on

Tal Boss

Tal is the app where bosses meet engineers & builders directly, without middlemen

Download Tal Boss

Sakshi Tantak on Tal

Headline

Senior Data Scientist, Acko | Deep Learning | topmate.io/sakshitantak

About

I’m passionate about solving real-world problems with impactful, scalable tech. With hands-on experience across diverse datasets and industries, I help companies harness the power of AI to transform operations and design more human-centered, intelligent systems. My strength lies in translating complex use cases into practical AI solutions — driving value while advocating for a balanced relationship between humans and machines in the workplace. Outside of work, I explore life like a curious node in a vast neural network — swimming, meditating, singing, traveling, and observing the little things that make life beautifully insignificant. You might find me deep in research papers one day, and decompressing in a quiet park the next. Let’s talk AI, ML, DL — or just share notes on building a more mindful, tech-driven world. And if you needed a sign to move your body today… this is it. Always excited to collaborate with mission-driven teams shaping the future of safe, responsible AI 🫶🏻

Location

Bengaluru

AI stack

hugging face

langchain

langgraph

openai

Work experience

Data Scientist 2 at Acko, Mar 2024 - Present · 2 yrs 2 mos, Bengaluru, Karnataka, India, Built platform for identifying document fraud at the time of claim Built central platform that enables building any document intelligence use cases with the help of a mere configuration file. Conducting continued pre training and fine-tuning of SLMs for domain adaptation in insurance like analysing policies, and enabling JSON data interpretation using LLMs.

Senior Data Scientist at Vakilsearch, Jun 2023 - Mar 2024 · 10 mos, India, Conducted continued pre-training and fine-tuning of language models ranging from older to relatively newer architectures to serve and enhance legal domain. Built a chatbot to help reduce escalations in customer experience section with LLMs using RAG

Senior Data Scientist at i3systems.ai, May 2022 - Jun 2023 · 1 yr 2 mos, Mumbai, Maharashtra, India, - Worked on developing models and strategising data for Financial Underwriting platform - an intelligent platform to extract and validate data from financial documents like Salary Slips & ITR. Trained CRF-BiLSTM based architectures using flairnlp & CRF++ for NER with a very generic tagging strategy for documents where data exists in key-value pairs - Fine-tuned Clinical LongFormer for extraction of clinical & non-clinical data from discharge summaries - Fine-tuned LiLT for extraction of major sections from discharge summaries - Conducted trial to predict quality of document images using LayoutLMv3 - Fine-tuned LayoutLM, LayoutLMv2 & LiLT for extraction of PII from KYC documents - Aadhar, PAN Card, Passport, Voter ID & Driving Licence - Conducted trial for image classification by fine-tuning Vision Transformer based architectures - Worked on correcting angles of document images to prevent issues arising from disoriented images in inference of multimodal models - Built quick demos and internal annotation tools using Streamlit - Customised existing streamlit component for NER annotation/highlighting to aid highlighting entire token instead of substrings selected upon dragging - Customised existing streamlit component to convert BIO to BIOES style of annotation for NER - Working on developing a system to identify potentially fraudulent claims by using a vast array of data points extracted from documents attached with claims, using objective triggers & embeddings based search & retrieval of claims with similar to exact section occurrence in documents

Data Scientist at i3systems.ai, Aug 2021 - May 2022 · 10 mos, Mumbai, Maharashtra, India, - Trained NER models to extract Metadata from Bills, Reports & Discharge Summaries using flairnlp, SpaCy & CRF++ - Integrated Microsoft Azure’s Language Services in an existing product for medical codes like ICD10, UMLS, etc using textanalytics healthcare API - Trained NER models for extraction of PII information from KYC documents - Aadhar, PAN Card, Passport, Voter ID, Driving Licence & Insurance IDs - Experimented with a lot of models for multiple tasks

Python Developer at Indian Institute of Technology, Bombay, May 2021 - Sep 2021 · 5 mos, India

Junior ML Engineer at Omdena, May 2021 - Jul 2021 · 3 mos, India, NLP, Web Scraping

Education

Bachelor of Engineering, Information Technology, Pune Institute of Computer Technology

Public links

github: https://github.com/pikaduck