Laalini B
Data Scientist @ CBA |Staff ML Systems Engineer | inference optimization & JAX/PyTorch Pretraining | Ex-JPMC | MBA (IIM-K) | Author & Patent Holder
Bengaluru
12+ years of work experience
Feb 20267 mosPresent
Multi-Agent Inference: Architectured multi-agent reasoning pipelines using Google ADK by using intent recognition and deterministic workflow orchestration that calls multiple tools and agents depending upon the intent identified
Evaluation Engineering: Wrote AI evals to assess agentic pipeline quality end-to-end, including custom judges and custom eval suites, and integrated them into observability platforms such as HoneyHive for continuous monitoring.
Platform Scalability : Scaled the virtual relationship manager for 250 thousand business banking users and monitored constantly for the model responses and client feedback regarding the interface and added additional intents and fallback mechanisms to support it
Model Governance : Worked on PII Gaurdrails , JailBreak Gaurdrails , Vulnerability Detection Gaurdrails, and compliance guardrail to ensure secure and user-friendly chat that is approved by regulatory authorities and secure enough for a bank
Scientific Accountability: Translated research papers on the jailbreaking to create the jailbreak gaurdrail
Jan 20251 yr 1 moJan 2026
Multi-Agent Inference: Designing multi-agent reasoning systems over structured enterprise data, accelerating insight extraction by ~30% across production workflows.
Evaluation Engineering: Building evaluation and error-analysis pipelines, improving retrieval accuracy, reliability, and scientific traceability in live GenAI systems.
Platform Scalability: Operating distributed inference services on AWS, supporting 50+ internal teams while maintaining consistent 99.9% uptime.
Model Governance: Establishing controlled evaluation baselines to validate GenAI behavior changes across prompt tuning, architecture updates, and data refresh cycles.
AI Authenticity: Integrating LLM-as-Judge and RAG
Feb 20241 yrJan 2025
Greater Bengaluru Area
Collaborated with stakeholders and product owners to define problem statements and developed a Gen AI-based technical architecture to address them, guiding the project from concept to deployment.
Supervised a cross-functional team, including ETL, Backend, and Research teams, resulting in a 20% reduction in project turnaround time and a successful product launch.
Led the development of a Generative AI-based Plant Maintenance Copilot, optimizing accuracy from 50% to 70%through innovative techniques such as Tree RAG and advanced metadata creation.
Evaluated and implemented Large Language Models (LLMs) using the RAAGAS framework, pioneering the development of an LLM agent reflection-based evaluation framework to enhance model performance.
Designed and deployed an ETL pipeline for the Maintenance Assist Co-Pilot using Azure Cognitive Services, including Azure Form Recognizer, Azure AI Search, and Azure OpenAI embeddings. Ensured reliable performance with 99% uptime by deploying in Azure Kubernetes Service.
Established AI/ML best practices and standards, integrating security and vulnerability management into AI/ML pipelines and ensuring compliance with regulatory requirements and industry standards.
Apr 202311 mosFeb 2024
*NLP Engineer (Project 2: Automated Customer Support)*
- Developed a state-of-the-art automated customer support system using Transformers and BERT models.
- Designed and implemented conversational AI solutions for enhancing customer interactions.
- Leverage NLP techniques to understand and respond to customer queries effectively.
- Collaborated with cross-functional teams to integrate the solution into the company's customer support platform.
Utilized RAG and fine tuning on top of the LLama models for this
*NLP Engineer (Project 1: Email Automation for Recruiters)*
- Implemented email automation for recruiters (Roberthalf) using LLMs.
- Generated automated customized emails for recruiters to send to candidates using Flan-T5 models.
- Worked on Fine tuning LLama models to perform the above tasks
Feb 20185 yrs 2 mosMar 2023
Hyderabad, Telangana, India
NLP Engineer / Machine Learning Engineer (02/2018-01/2022)
Project Clients: TIAA, Manpower Experis, RBS, Nokia MLOPs
Project 1 : Auto recommendation of candidates for Job order(SPACY ) : Manpower Xperis -The goal of this project is to fully automate the hiring process, from obtaining job descriptions from employers to locating qualified candidates through a variety of job portals and distributing the candidate summary document.
The entire project is developed on SPACY Framework
Identified the most relevant candidate resumes from different job portals and internal databases and saved 80% of manual extraction time
Extracted the entries like skills, location, experience and job title from the job description using spacy models with an accuracy of 90%
Used NLP techniques to work on similarity methods for learning the mapping between job descriptions and resumes
Reduced processing time by up to 80% for each job description by automating the process of obtaining job descriptions from companies using Outlook.
Designed a Web Scraper utility using Beautiful Soup with Python technologies to scrape open-source public candidates’ data/resumes for NLP tasks.
Project 2 : RBS Autoclosure
Worked on classifying high risk transaction alerts from a huge dataset of alert summaries where I used the LGBM algorithm and also implemented Bayesian optimization to effectively tune the hyperparameters.
Saved 3.5 million pounds for RBS as part of this project.
Aug 20143 y 5 moJan 2018
Hyderabad
Laalini’s AI Stack
Education
Master of Business Administration - MBA, Business Administration and Management, General
Indian Institute of Management, Kozhikode
Bachelor of Technology (B.Tech.), Electrical and Electronics Engineering
Jntu kakinada
20104 yrs2014
Join Laalini and 10k+ other professionals on
Tal is the app where bosses meet engineers & builders directly, without middlemen
Download Tal Boss