Dr. Phani Siginamsetty
Associate Data Scientist · Hexaware Technologies · Chennai, India

Dr. Phani
Siginamsetty

I build Autonomous AI Agents

PhD-qualified Data Scientist and AI Engineer with 5+ years bridging cutting-edge academic research with enterprise-grade production systems. Expert in designing autonomous Multi-Agent Systems — where LLM-powered agents plan, reason, use tools, and collaborate to solve complex real-world workflows end-to-end. Deep hands-on experience across the full AI lifecycle: from RAG pipeline architecture and LLM fine-tuning (PEFT/QLoRA) to quantized edge deployment, MLOps, and scalable cloud infrastructure on AWS. Proven ability to translate research breakthroughs in multilingual NLP, multimodal AI, and quantum-motivated algorithms into production-ready systems — backed by 11 patents and 7 peer-reviewed publications in Elsevier, IEEE, and Springer. Passionate about building intelligent systems that are not just accurate, but autonomous, explainable, and deployable at scale.

Agentic AI Multi-Agent Systems Generative AI RAG Pipelines LLM Fine-Tuning Data Science Computer Vision NLP Research MLOps · AWS Edge AI LangGraph · CrewAI Vector Databases AI Research 11 Patents 7 Publications
View Projects Resume Contact
Expertise Distribution
62+ Skills
0
Patents Filed
0
Granted
0
Publications
9.25
PhD CGPA
0
Yrs Experience
0
Key Projects
Technical Expertise

Skills & Stack

Full-spectrum ML/AI stack — hover bars, click categories to explore.

Core Proficiency
Generative AI & LLMs95%
Multi-Agent Systems92%
RAG & Vector Search90%
Python & FastAPI88%
Data Science & ML85%
AWS Cloud & MLOps83%
Computer Vision80%
NLP Research78%
Domain Focus
Technology Stack
GenAI & Agentic Frameworks
11 technologies
CrewAIAutoGenLangGraphAgnoPydanticAIHaystackTool-CallingSemantic RoutingHITLRagasTruLens
LLMs & Model Engineering
12 technologies
AWS BedrockLlama 3.2GPT-4oPEFTQLoRAUnslothDPO/RLHFDeepSpeedFSDPGGUFAWQGPTQ
Data Science & Advanced ML
9 technologies
PyTorchScikit-LearnRL (PPO, DQN)Time-SeriesXGBoostRandom ForestSiamese NetworksAutoMLQuantum-Motivated Algorithms
Vision & Multimodal AI
8 technologies
Multimodal RAGOpenCVAWS TextractDocument IntelligenceWhisper (ASR)ElevenLabs (TTS)CLIPImage & Audio Processing
MLOps, Cloud & Backend
11 technologies
PythonAWS SageMakerAWS LambdaAWS EC2DockerKubernetesCI/CDFastAPIFlaskJWT/RBACOptuna
Data Infrastructure & Vector DBs
11 technologies
PineconeMilvusWeaviateChromaFAISSPostgreSQLMongoDBSparkPandasNumPyParquet
Career

Work Experience

5+ years across industry research, enterprise AI, and academia.

5+Years
5Roles
4Companies
2020Started
NowPresent
Associate Data Scientist
Hexaware Technologies · Chennai, India
AgnoMulti-AgentRAGLangGraphHITL
Mar 2025 – Present ● Current
  • Autonomous Fraud Detection: Spearheading real-time fraud detection using stateful multi-agent systems via the Agno framework with advanced tool-calling for complex transaction analysis.
  • Advanced RAG Pipelines: Engineering a multi-agent RAG pipeline for "Smart Tutor" using vector databases and semantic routing to deliver personalized content while minimizing hallucinations.
  • Enterprise Automation: Designing agentic workflows with LangChain and LangGraph to automate reporting with Human-in-the-Loop (HITL) mechanisms, reducing operational overhead.
Research Assistant
Volvo Group · Bangalore, India
Edge AIGGUF/AWQComputer VisionCNN
Jun 2024 – Mar 2025
  • Edge GenAI & Quantization: Researched lightweight LLMs for on-device inference using GGUF/AWQ quantization to reduce memory footprint and latency on vehicular hardware.
  • Computer Vision Diagnostics: Deployed an optimized CNN pipeline for real-time component recognition within the Vehicle Configuration Manager (VCM) to automate visual inspections.
Data Science Researcher (PhD Scholar)
SRM University AP · Amaravati, India
RAGFastAPImT5NLPQuantum AIPatents
Sep 2021 – Jul 2024
  • Healthcare AI: Architected a secure RAG chatbot for SRM Global Hospital to retrieve medical protocols while ensuring strict data privacy and proprietary data embedding.
  • Audio Intelligence: Engineered a MoM automation API using FastAPI, STT, and speaker diarization to autonomously extract abstractive summaries and action items from recordings.
  • Multilingual NLP: Developed MATSFT and MMSFT frameworks by fine-tuning mT5 for low-resource Indian languages, resulting in multiple high-impact journal publications.
  • Quantum AI & IP: Architected quantum-motivated summarization processors for data compression, leading to multiple Indian Patents including 1 Granted Patent.
Trainee Engineer (ML Research)
Tychee Innovations · Andhra Pradesh, India
Object DetectionMLHealthcare AIPredictive Analytics
Aug 2020 – Jul 2021
  • Industrial Safety Vision: Deployed real-time object detection to monitor hazardous machinery, triggering emergency stops via spatial tracking of hand proximity to danger zones.
  • Predictive Analytics: Developed ML models to forecast patient outcomes and translate clinical data into actionable insights for data-driven healthcare decisions.
Assistant Professor
Dhanekula Engineering College · Andhra Pradesh, India
PythonDSAMentorshipTeaching
Oct 2020 – Aug 2021
  • Software Mentorship: Instructed Data Structures, Algorithms, and Python, mentoring students in software engineering best practices and technical problem-solving.
Portfolio

Key Projects

Enterprise-grade AI systems built end-to-end — spanning GenAI, fraud detection, computer vision, and healthcare.

Tech Stack Distribution
4Projects
Domain Coverage
AI Techniques Used
01 · SmartTutor
AI-Powered Knowledge Base & Interactive Learning Platform
FastAPIAWS BedrockAmazon Nova ProClaude Sonnet 4PineconeAWS PollyPostgreSQLAWS S3JWT/RBAC
Document Processing
  • Multi-format: PDF, DOCX, PPTX, PNG, JPG
  • Auto-conversion of Word & PowerPoint to PDF
  • Intelligent image extraction & spatial analysis
  • Custom prompt instructions per document
  • Module regeneration with new instructions
AI-Powered Features
  • Structured module generation via PHASE 1–3 analysis
  • Interactive AI tutor with AWS Polly TTS (4 voices)
  • Rich image explanations with analogies
  • Auto question generation & answer validation
  • Internet search integration for extended context
Knowledge Management
  • Vector search via Pinecone + Titan Embed (1024-dim)
  • Document sharing with edit proposals & review
  • Course publishing for trainee access
  • Real-time progress tracking during processing
  • Markdown-supported module curation
Multi-User & RBAC
  • Roles: Super Admin, Trainer, Trainee
  • Process-based org grouping (departments)
  • Admin approval workflow for new trainers
  • Configurable rate limits per role (SlowAPI)
  • JWT auth with bcrypt password hashing
Interactive Learning
  • TTS lectures with pause-for-questions
  • Raise Hand feature during live sessions
  • Voice input via Google Speech Recognition
  • Quiz system with instant AI feedback
  • Preview mode for trainers before publishing
Security & Infra
  • AWS RDS PostgreSQL + SQLAlchemy ORM
  • S3 server-side AES256 encryption
  • CORS + SQL injection protection
  • Async operations via asyncio
  • Rotating file handler logging
Full Tech Stack
Python 3.9+ FastAPI PostgreSQL (AWS RDS) SQLAlchemy Pinecone AWS Bedrock Amazon Nova Pro v1 Claude Sonnet 4 Titan Embed v2 (1024-dim) AWS Polly AWS S3 PyMuPDF Google Speech API JWT bcrypt SlowAPI asyncio Showdown.js
02 · Fraud Prevention
Enterprise Fraud Prevention System — Citi Bank
PythonXGBoostMulti-AgentAWS
  • Hybrid Risk Engine: Dual-layered system fusing statistical anomaly detection (XGBoost) with GenAI-driven forensics, reducing investigation time and false positives.
  • Autonomous Rule Discovery: Multi-agent workflow for live transaction monitoring, detecting zero-day fraud patterns with Human-in-the-Loop oversight.
  • Argus Agent: AI Data Analyst using msoffcrypto to securely decrypt and parse sensitive financial datasets locally for evidence-based risk verdicts.
03 · Cheque Verification
Automated Bank Cheque Verification System
PyTorchAWS TextractOpenCVSiamese NN
  • Forensic Digitization: End-to-end vision pipeline using AWS Textract and OpenCV for layout analysis and digitization of MICR codes and payee details with high OCR accuracy.
  • Signature Verification: PyTorch-based Siamese Neural Network for one-shot learning, using contrastive loss and feature embeddings to detect forged signatures.
  • Cross-Modal Logic: NLP algorithms cross-verifying extracted semantic data (numeric vs. written amounts) to flag discrepancies for manual review.
04 · Medical AI
Personalized Medical AI Assistant
Llama 3AgnoLangGraphMongoDB
  • Clinical Guardrails & RAG: Domain-specific agent using Llama 3 and Vector DBs, with query expansion and re-ranking to ground answers exclusively in verified medical literature.
  • Stateful Memory: Persistent context-retention engine using LangGraph and MongoDB to map longitudinal symptoms and medical history for personalized health insights.
Intellectual Property

Research & Patents

11 patents filed · 3 granted · 7 peer-reviewed publications in Elsevier, IEEE & Springer.

0
Patents Filed
0
Granted
0
Publications
2022
First Patent
Patent Status
11 Total
Publications by Venue
Patents by Year
System and a method for automated exam evaluation and personalized learning feedback
202541018210 2025 Education AI
A System and a Method for Managing API Calls in A Large Language Model
202441096836 2024 LLM
A System and a Method for Healthcare Data Processing and Decision Support
202441076761 2024 Healthcare
System and method for multilingual fake news detection in multimodal information
202441030030 2024 NLP
A Healthcare Summarization System and A Method Thereof
202441005845 2024 Healthcare
A System and a Method for Personalized E-Content Generation Based on Student Performance
202441003347 2024 Education AI
System and method for deriving multilingual meeting minutes
Granted
202441001022 Grant: 581292 2024 NLP
System and method for multimodal multilingual input summarization using quantum motivated processors
Granted
202341005519 Grant: 66614 2023 Multimodal AI
A System and A Method for Generating Trading Coupons
202341007665 2023 FinTech
A System and A Method for Prediction of The Strength of Concrete
Granted
202341007257 Grant: 582851 2023 Engineering
A System and Method for Performing Multilingual Multimodal Summarization
202241073648 2022 Multimodal AI
[1]
MATSFT: User query-based multilingual abstractive text summarization for low resource Indian languages by fine-tuning mT5
Alexandria Engineering Journal · Elsevier 2025 Journal
Phani, S., et al.
10.1016/j.aej.2025.04.031
[2]
Improving Preliminary Clinical Diagnosis Accuracy through Knowledge Filtering Techniques in Consultation Dialogues
Computer Methods and Programs in Biomedicine · Elsevier 2024 Journal
Abdul, A., Phani, S., et al.
10.1016/j.cmpb.2024.108051
[3]
MMSFT: Multilingual Multimodal Summarization by Fine-tuning Transformers
IEEE Access 2024 IEEE
Phani, S., et al.
10.1109/ACCESS.2024.3454382
[4]
MMSML: Multilingual Multimodal Summarization for Multimodal Input
Intl. Conference on Data Science and Applications · Springer 2024 Conference
Phani, S., et al.
10.1007/978-981-96-2724-0_5
[5]
Recognition for Attendance System Using Reinforcement Learning
FICTA · Springer 2023 Conference
Phani, S., et al.
10.1007/978-981-99-6702-5_15
[6]
Abstractive Text Summarization with Fine-Tuned Transformer
MAI 2022 · Springer 2023 Conference
Phani, S., et al.
10.1007/978-981-99-0189-0_46
[7]
Machine Learning Classifiers and TPOT Classifier (AutoML) to Predict Readmission Patterns of Diabetic Patients
IJRTE 2020 Journal
Phani, S., et al.
10.35940/ijrte.f7415.059120
Academic Background

Education

A decade of academic excellence from SSC through PhD.

2021 – 2025
Ph.D.
Computer Science & Engineering
SRM University AP
9.25 / 10.0
Amaravati, India
11 Patents · 7 Publications
Multilingual NLP, Quantum AI, RAG
2018 – 2020
M.Tech
Computer Science & Engineering
KL University
8.5 / 10.0
Andhra Pradesh, India
Advanced ML & Deep Learning
2014 – 2018
B.Tech
Computer Science & Engineering
JNTUK
81.0%
Andhra Pradesh, India
CS Fundamentals, Algorithms, Networks
2012 – 2014
Intermediate
MPC
Board of Intermediate Education
91.70%
Andhra Pradesh, India
Maths, Physics, Chemistry
2011 – 2012
SSC
10th Standard
Board of Secondary Education
9.3 / 10.0
Andhra Pradesh, India
Academic Foundation
Document

Resume

Download PDF