Agentic AI Developer, UST Global

Asirwad Sali

Agentic AI Developer

Specialized Agentic AI Developer with expertise in building intelligent, autonomous systems using LangChain, LangGraph, and enterprise AI frameworks. Passionate about creating secure, scalable agentic solutions that transform healthcare and enterprise workflows.

Connect with me:
Scroll down

About Me

I'm a dedicated Agentic AI Developer with specialized expertise in building intelligent, autonomous systems that leverage multiple AI agents to solve complex business problems. Currently working at UST Global in the US Healthcare sector, I design and implement enterprise-grade agentic AI systems using cutting-edge frameworks like LangChain, LangGraph.

My core competencies lie in developing multi-agent orchestration systems, secure PHI tokenization workflows, and robust backend architectures using Spring Boot microservices. I've successfully architected clinical assistant platforms that automate intake processes, ambient scribing, and structured SOAP note generation, while ensuring HIPAA compliance and data security.

With a strong foundation in both traditional software engineering and emerging AI technologies, I bridge the gap between theoretical AI research and practical enterprise applications. My work focuses on creating reliable, scalable agentic systems that can reason, plan, and execute complex workflows with minimal human intervention. I'm particularly passionate about applying agentic AI to transform healthcare delivery and improve patient outcomes.

8.5
CGPA
2024
Graduate

Education

Bachelor of Technology in Computer Science and Engineering

College of Engineering Cherthala
2020-24
CGPA: 8.5

Class 12 - Computer Science

SNTHSS Cherthala
2018-20
Percentage: 94.10%

Experience

A timeline of professional growth and technical achievements

Agentic AI Developer

UST Global - US Healthcare
Aug 2024 – Present
Full-time

Specializing in developing enterprise-grade Agentic AI systems for US healthcare workflows using cutting-edge frameworks like LangChain, LangGraph and AWS Bedrock, Strands Agents.

~3x
Throughput improvement
HIPAA
Compliant systems
Multi-agent
Orchestration

Key Achievements

  • Developing enterprise-grade Agentic AI systems for US healthcare workflows using LangChain, LangGrap and AWS Strands with focus on HIPAA, PHI tokenization, audit trails and traceability
  • Designed and implemented multi-agent systems (scribe, intake, reasoning, validation) enabling tool-use planning, stepwise decomposition and workflow orchestration for clinical tasks
  • Built RAG pipelines and retrieval systems using vector embeddings, chunking, and BM25/OpenSearch fallbacks for accurate clinical summarization, medication and entity extraction
  • Engineered secure PHI tokenization and redaction flows with HashiCorp Vault Transit Engine + Spring Boot microservices integrated into orchestration layers to ensure data minimization and auditability
  • Integrated LLMs with safety and grounding layers (Bedrock, LLaMA, Claude, Gemini) with prompt/version control, hallucination mitigations, JSON output schemas and downstream validation
  • Developed LLMOps and reliability tooling: prompt versioning, prompt/eval pipelines (RAGAS-style), observability, telemetry, retry supervisors, and rate-limiting for production stability
  • Improved backend throughput and scalability: moved tokenization and async processing to cross-account AWS Lambda and managed async queues, improving end-to-end throughput by ~3x
  • Practiced enterprise engineering: CI/CD (GitHub Actions), containerization (Docker), infra-as-code patterns, sprint-based delivery, code reviews, and post-mortems

Technologies Used

LangChainLangGraphDSPyGenKitAWS BedrockSpring BootHashiCorp VaultAWS LambdaS3STSDockerGitHub ActionsPostgreSQLWeaviate
Continuing to build innovative solutions...

Selected Projects

Innovative solutions spanning GenAI, healthcare technology, and deep learning

MedScribeAI – GenAI Clinical Assistant Platform

Award
Healthcare Technology
2025
Runner-Up – UST Agentic AI Hackathon
Runner-Up – UST Agentic AI Hackathon 2025

An innovative multi-agent, HIPAA-compliant clinical assistant platform that leverages ambient conversation scribing to revolutionize healthcare documentation.

2nd Place
Team Ranking
70+
Competing Teams
Runner-Up
Recognition

Key Highlights

  • Spearheaded the architecture and full-stack development of a multi-agent, HIPAA-compliant clinical assistant with ambient conversation scribing capabilities using Next.js, GenKit, Spring Boot, Firebase, and LLaMA 4.0
  • Engineered a secure PHI tokenization system using HashiCorp Vault Transit Engine with SpringBoot REST microservices, ensuring zero-trust data storage and compliance
  • Implemented Google GenKit agents featuring contextual chat interfaces, automated patient intake processes, and EHR-ready clinical documentation pipelines
  • Developed a comprehensive API layer with advanced redaction capabilities to ensure patient privacy, auditability, and GenAI safety in healthcare workflows
  • Successfully delivered a production-ready demonstration to stakeholders through full-stack deployment, earning recognition as a finalist among 70+ competing teams

Technologies

Next.jsGoogle GenKitSpring BootFirebaseLLaMA 4.0HashiCorp VaultHIPAA ComplianceGenerative AI

ProactivePulseAI – Agentic AI for MSPs

Award
Managed Service Providers
2025
Selected for Round 1
Selected – ProactivePulseAI

An agentic platform that correlates helpdesk ticket trends with infrastructure metrics to surface root-cause hypotheses, actionable remediation steps and prioritized alerts.

Round 1
Selection Status
MSPs
Platform Focus

Key Highlights

  • Contributed to key multi-agent modules: ingestion, anomaly detection, ticket analysis, correlation engine, insight generator and notification pipelines
  • Implemented production patterns: DynamoDB with GSIs for low-latency queries, AWS OpenSearch for full-text and fuzzy search
  • Built APIs using FastAPI and dashboards using NextJS for paginated insights, on-demand multi-agent runs and historical investigations
  • Emphasized observability and audit trails throughout the platform architecture

Technologies

LangChainLangGraphAWS BedrockFastAPINext.jsDynamoDBOpenSearchBedrock-powered NLP

Qualen – AI-Powered Test Case Generation for Healthcare QA

Award
Quality Assurance
2025
Team Lead
Qualen – Generate compliant, auditable test cases

An LLM-first QA copilot that converts requirements into positive/negative/boundary test suites with end-to-end traceability.

Healthcare QA
Focus Area
Traceability
Key Features

Key Highlights

  • Delivered HIPAA-conscious compliance features: exportable evidence, explainable test mapping, status badges and an always-on traceability matrix to support audits and regulatory reviews
  • Built LLM-driven workflows using Google Gemini for test generation, conversational QA copilot UX, and ALM integrations (Jira / Azure DevOps / Polarion)
  • Implemented scalable analytics and storage with BigQuery for traceability insights, serverless generation pipelines (Cloud Run / Vercel)
  • Enabled event-driven sync to ALM systems for continuous integration

Technologies

Next.js 15TypeScriptshadcn/uiTailwindBigQueryGemini APIALM integrationsVercelCloud Run

Privacy-Safe Agentic Clinical Triage Assistant (LCAC Framework)

Award
Healthcare Security
2025
Contributor
LCAC Framework Implementation

An agentic clinical triage assistant implementing the Least-Context Access Control (LCAC) framework for cognitive zero-trust in clinical AI.

LCAC
Security Model
Clinical Triage
Application

Key Highlights

  • Contributed to implementation of the Least-Context Access Control (LCAC) framework for cognitive zero-trust in clinical AI
  • Focused on privacy-safe agentic workflows that minimize data exposure while maintaining clinical utility
  • Implemented secure context management for multi-agent clinical decision support systems
  • Ensured compliance with healthcare data protection regulations

Technologies

LangChainLangGraphAWS BedrockSpring BootHashiCorp VaultHIPAA Compliance

InvisiCipher++ – GAN-based Steganography Suite

Award
Cybersecurity & AI
12 Months
B.Tech Major Project (Team Lead)
B.Tech Major Project

An advanced steganography system that combines cutting-edge generative adversarial networks for ultra-realistic image synthesis and secure data embedding.

92%
PSNR Achievement
50%
Error Reduction
-30%
Time Optimization

Key Highlights

  • Architected a novel hybrid image processing pipeline integrating StackGAN and ESRGAN architectures for generating ultra-realistic cover images
  • Engineered a sophisticated CNN-based encoder/decoder system (STEGOCNN) that achieved 92% PSNR and reduced steganographic embedding error by 50%
  • Implemented advanced algorithmic optimizations including bit-plane slicing and adaptive quantization techniques, reducing model training time by 30%
  • Developed a comprehensive full-stack desktop application using PyQT5 and created an interactive web-based demonstration using Streamlit for stakeholder presentations

Technologies

StackGANESRGANConvolutional Neural NetworksPyTorchPyQT5StreamlitDeep LearningComputer Vision

Skills & Recognition

Technical expertise and professional achievements

Technical Expertise

Programming Languages

JavaPythonTypeScriptJavaScriptSQL

Frameworks & Libraries

Spring BootLangChainLangGraphDSPyGenKitDjangoHibernateReactNext.js

Cloud & DevOps

Amazon Web Services (AWS)AWS BedrockAWS LambdaAWS S3AWS IAMDockerGitHub ActionsCI/CD PipelinesMonitoringHashiCorp Vault

Databases & Storage

PostgreSQLMySQLAmazon DynamoDBMongoDBWeaviateRedis

AI & Methodologies

Agentic AIMulti-Agent SystemsRAG PipelinesLLMOpsPrompt EngineeringHIPAA ComplianceAgile/ScrumTest-Driven Development

Computer Science Fundamentals

Data Structures & AlgorithmsOperating SystemsConcurrencyObject-Oriented ProgrammingSystem DesignDesign Patterns

Areas of Interest

Scalable Backend Systems
Deep Learning
Generative AI
Secure Cloud Architectures

Project Management

Agile
Jira
Sprint Planning
Code Reviews
Post-Mortems

Soft Skills

Ownership
Problem Solving
Communication
Leadership
Adaptability

Certifications

Certified Full Stack Java Developer

Global Skill Development Council (GDSC)3 Months

Comprehensive certification covering Data Structures & Algorithms, Advanced Spring Boot, REST APIs, Microservices Architecture, Hibernate, and Frontend Integration. Demonstrates proficiency in building scalable, enterprise-grade applications.

Problem Solving Using Computational Thinking

University of Michigan (Coursera)1 Month

Applied Computational Thinking concepts to real-world problems with systematic and algorithmic approaches. Enhanced analytical and structured problem-solving abilities.

Achievements

Recognition for innovation and technical excellence

Runner-Up – UST Agentic AI Hackathon

Competition
2025

Secured 2nd Place

Project: MedScribeAI – GenAI Clinical Assistant Platform

Led the development of a groundbreaking multi-agent, HIPAA-compliant clinical assistant platform that leveraged ambient conversation scribing to revolutionize healthcare documentation.

Key Highlights
Secured 2nd place among 70+ competing teams in a prestigious company-wide hackathon
Delivered a full-stack Generative AI healthcare solution with real-world applicability
Implemented HIPAA-compliant architecture ensuring patient data security and privacy
Presented solution to executive stakeholders and technical panel with live demonstration
Technologies
Next.jsGoogle GenKitSpring BootFirebaseLLaMA 4.0HashiCorp Vault
Featured

Selected – ProactivePulseAI (Agentic AI for MSPs)

Project Contribution
2025

Round 1 Selection

Project: ProactivePulseAI – Agentic AI for MSPs

Contributed to ProactivePulseAI: an agentic platform that correlates helpdesk ticket trends with infrastructure metrics to surface root-cause hypotheses, actionable remediation steps and prioritized alerts.

Key Highlights
Selected in Round 1 for contribution to key multi-agent modules: ingestion, anomaly detection, ticket analysis, correlation engine, insight generator and notification pipelines
Implemented production patterns: DynamoDB with GSIs for low-latency queries, AWS OpenSearch for full-text and fuzzy search
Built APIs using FastAPI and dashboards using NextJS for paginated insights, on-demand multi-agent runs and historical investigations
Emphasized observability and audit trails throughout the platform architecture
Technologies
LangChainLangGraphAWS BedrockFastAPINext.jsDynamoDBOpenSearch
Featured

B.Tech Major Project Team Lead

Academic
2024

Academic Excellence

Project: InvisiCipher++ – GAN-based Steganography Suite

Spearheaded a team to develop an innovative steganography system combining cutting-edge generative adversarial networks for ultra-realistic image synthesis and secure data embedding.

Key Highlights
Demonstrated exceptional leadership guiding a team of 4 through the complete project lifecycle
Achieved outstanding 92% Peak Signal-to-Noise Ratio (PSNR) for image quality preservation
Reduced steganographic embedding error by 50% through advanced algorithmic techniques
Optimized model training efficiency by 30% using algorithmic optimizations
Technologies
StackGANESRGANConvolutional Neural NetworksPyTorchPyQT5Streamlit
2nd
Hackathon Place
70+
Competing Teams
2025
Recent Awards

Get In Touch

Ready to collaborate? Let's create something amazing together

Let's Connect

I'm always open to discussing new opportunities, innovative projects, or just having a conversation about technology and development.

Location

India (Remote)

Follow Me

I'll get back to you as soon as possible. Usually within 24 hours.

Currently viewing: hero section