Ankit Jaiswal profile photo 1 on Tal

Ankit Jaiswal

Sr Data Scientist at JIO Platforms Ltd. | Ex- Barclays | PGDBA @ IIM Calcutta, ISI Kolkata, IIT Kharagpur | Btech @ NIT Allahabad

Bengaluru

11+ years of work experience

Jio Platforms Limited (JPL)
Dec 20223 yrs 9 mosPresent

Bengaluru

Led an LLM & RAG–based, template-agnostic entity extraction platform for complex invoices and contracts, designing a Retrieval-Augmented Generation pipeline with PaddleOCR for table/text segmentation, the Unstructured library for OCR normalization, and pre-trained LLMs for robust multi-template entity extraction.
AppZen
Jul 20211 yr 6 mosDec 2022

Pune, Maharashtra, India

AppZen
Jul 20211 yr 5 mosNov 2022

Led design of a CNN-autoencoder anomaly detection pipeline for authentication model, reducing audit latency and filtering noisy false positives/negatives for finance auditors. Developed an end-to-end receipt classification service for Chinese receipts using ResNet-50, OCR post-processing, and Protobuf serialization to AWS S3, with robust retry/backoff for ingestion failures. Acted as an L3 tech lead, driving faster closure of customer tickets related to model outputs, creating debugging playbooks, and collaborating closely with product/support. Instituted model quality gates—PR reviews, regression test suites, semantic versioning, and coordinated rollouts with Release Engineering—which significantly reduced production incidents. Integrated explainability modules into production ML pipelines, enabling finance users to transparently interpret model predictions and facilitating audit compliance by surfacing feature attributions within the user workflow.
Barclays UK
May 20192 yrs 3 mosJul 2021

Pune/Pimpri-Chinchwad Area

Built an Adverse Media Detection system as part of KYC, combining spaCy NER/POS with rules, spell correction, coreference, and negation handling; achieved 98% precision on validation and exposed via Flask + Dash UI for analysts. Developed a contextual search / QA engine used across business units: BM25 + SIF-weighted embeddings plus transformer models (BERT, GPT-2) to re-rank results, improving query latency (~70% faster) and NDCG@5 by 20% vs legacy search. Designed and deployed a Rasa-based chatbot, integrated with Facebook Messenger and Rasa X UI; reached 92% intent accuracy and 95% entity extraction across four conversational competencies. Created a complaints NLP pipeline using LDA2Vec topic modeling to cluster root causes and validate labels, driving 40% overall and up to 72% reduction in complaints for targeted segments. Implemented a Spark-based ALS recommender to nudge customers toward digital journeys, increasing conversions to digital channels by 24% compared to the existing rule-based model. Built a custom OCR engine using CNNs for scanned documents (97.6% entity accuracy), deployed on OpenShift using Flask APIs; the solution saved an estimated 8–10 FTE worth of manual data entry.
Barclays UK
Oct 20188 mosMay 2019

HPCL-Mittal Energy Limited
Jun 20143 yrs 1 moJun 2017

India

Worked in Electrical Maintenance and Equipment Reliability Dept.

Ankit’s AI Stack

opencode
opencode
claudecode
claudecode

Education

Indian Institute of Management, Calcutta

PGDBA, Business Analytics and Data Science

Indian Institute of Management, Calcutta

20172 yrs2019
Indian Institute of Technology, Kharagpur

PGDBA, Business Analytics and Data Science

Indian Institute of Technology, Kharagpur

20172 yrs2019

Join Ankit and 10k+ other professionals on

Tal Boss

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

Download Tal Boss

Ankit Jaiswal on Tal

Headline

Sr Data Scientist at JIO Platforms Ltd. | Ex- Barclays | PGDBA @ IIM Calcutta, ISI Kolkata, IIT Kharagpur | Btech @ NIT Allahabad

Location

Bengaluru

AI stack

opencode

claudecode

Work experience

Senior Data Scientist at Jio Platforms Limited (JPL), Dec 2022 - Present · 3 yrs 9 mos, Bengaluru, Led an LLM & RAG–based, template-agnostic entity extraction platform for complex invoices and contracts, designing a Retrieval-Augmented Generation pipeline with PaddleOCR for table/text segmentation, the Unstructured library for OCR normalization, and pre-trained LLMs for robust multi-template entity extraction.

Sr. Machine Learning Developer at AppZen, Jul 2021 - Dec 2022 · 1 yr 6 mos, Pune, Maharashtra, India

Senior Machine Learning Engineer at AppZen, Jul 2021 - Nov 2022 · 1 yr 5 mos, Led design of a CNN-autoencoder anomaly detection pipeline for authentication model, reducing audit latency and filtering noisy false positives/negatives for finance auditors. Developed an end-to-end receipt classification service for Chinese receipts using ResNet-50, OCR post-processing, and Protobuf serialization to AWS S3, with robust retry/backoff for ingestion failures. Acted as an L3 tech lead, driving faster closure of customer tickets related to model outputs, creating debugging playbooks, and collaborating closely with product/support. Instituted model quality gates—PR reviews, regression test suites, semantic versioning, and coordinated rollouts with Release Engineering—which significantly reduced production incidents. Integrated explainability modules into production ML pipelines, enabling finance users to transparently interpret model predictions and facilitating audit compliance by surfacing feature attributions within the user workflow.

Machine Learning Developer/ Data Scientist at Barclays UK, May 2019 - Jul 2021 · 2 yrs 3 mos, Pune/Pimpri-Chinchwad Area, Built an Adverse Media Detection system as part of KYC, combining spaCy NER/POS with rules, spell correction, coreference, and negation handling; achieved 98% precision on validation and exposed via Flask + Dash UI for analysts. Developed a contextual search / QA engine used across business units: BM25 + SIF-weighted embeddings plus transformer models (BERT, GPT-2) to re-rank results, improving query latency (~70% faster) and NDCG@5 by 20% vs legacy search. Designed and deployed a Rasa-based chatbot, integrated with Facebook Messenger and Rasa X UI; reached 92% intent accuracy and 95% entity extraction across four conversational competencies. Created a complaints NLP pipeline using LDA2Vec topic modeling to cluster root causes and validate labels, driving 40% overall and up to 72% reduction in complaints for targeted segments. Implemented a Spark-based ALS recommender to nudge customers toward digital journeys, increasing conversions to digital channels by 24% compared to the existing rule-based model. Built a custom OCR engine using CNNs for scanned documents (97.6% entity accuracy), deployed on OpenShift using Flask APIs; the solution saved an estimated 8–10 FTE worth of manual data entry.

Data Scientist at Barclays UK, Oct 2018 - May 2019 · 8 mos

Engineer at HPCL-Mittal Energy Limited, Jun 2014 - Jun 2017 · 3 yrs 1 mo, India, Worked in Electrical Maintenance and Equipment Reliability Dept.

Education

PGDBA, Business Analytics and Data Science, Indian Institute of Management, Calcutta

PGDBA, Business Analytics and Data Science, Indian Institute of Technology, Kharagpur