Irving, TX // Status: Online // Open to Work

TANAY
TAMMINENI.

 

MS CS · 3.9 GPA · Published Researcher · AI Systems Developer Intern @ VoiceBotics AI
Bridging real-time deep learning performance with production-grade LLM systems.

DeploymentsResume ↗GitHub ↗
Currently building: AI-powered portfolio tools — live terminal, JD matcher, semantic project search
// About Me

From CV Research
to Production LLMs.

I started with computer vision — building real-time vehicle detection systems and publishing peer-reviewed research. That obsession with making AI work in the real world never left.

At VoiceBotics AI, I've shipped production RAG pipelines, real-time transcription systems, and end-to-end LLM fine-tuning infrastructure. I don't just prototype — I deploy.

I'm an MS grad from Southeast Missouri State (3.9 GPA), originally from Hyderabad, India, now based in Irving, TX. I bring a global perspective and a bias toward building things that actually work.

// What I Bring
⚡Production-Grade AI

I build LLM pipelines, RAG systems, and CV models that run in production — not just Jupyter notebooks. Deployed at VoiceBotics AI serving real enterprise clients.

📄Published Research

Peer-reviewed researcher with published work in computer vision. I understand the gap between academic AI and production AI — and how to bridge it.

🔧Full-Stack AI Engineering

From model fine-tuning and RAG architecture to FastAPI backends and React dashboards — I can own the entire AI product stack end to end.

MS GPA0.0
ResearchPeer-Reviewed
Projects0
Experience0+ Years
// Python
// LLMs
// RAG Pipelines
// Deep Learning
// Computer Vision
// Cryptography
// FastAPI
// Agentic AI
// DeepSpeed
// LangGraph
// Python
// LLMs
// RAG Pipelines
// Deep Learning
// Computer Vision
// Cryptography
// FastAPI
// Agentic AI
// DeepSpeed
// LangGraph

01.Architectures & Deployments

Click any card to flip and see full details

Feb 2026 — Mar 2026New

Distributed LLM Fine-Tuning Pipeline

Built distributed pipeline across 66 files and 11,000+ lines using QLoRA and DeepSpeed ZeRO-3 to fine-tune Llama 3 8B

PyTorchDeepSpeedQLoRAvLLMTerraformAWS

◈ Click to see details

Key Featuresclick to flip back

Distributed LLM Fine-Tuning Pipeline

  • ▸Built distributed pipeline across 66 files and 11,000+ lines using QLoRA and DeepSpeed ZeRO-3 to fine-tune Llama 3 8B
  • ▸Achieved 41.2% per-GPU memory reduction via MinHash LSH deduplication with 128 permutations across multi-GPU nodes
  • ▸Deployed with vLLM and AWQ quantization achieving 3.8x inference throughput improvement on Terraform-provisioned AWS
  • ▸Prometheus and Grafana monitoring tracking real-time GPU utilization across the training cluster
  • ▸Published findings in UniLLMOps covering QLoRA + DPO training, TIES/DARE/SLERP model merging, and vLLM deployment
PyTorch · DeepSpeed · QLoRA · vLLM · Terraform · AWS
GitHub ↗
Feb 2026 — Mar 2026New

Production RAG System

Built hybrid retrieval combining Qdrant dense search and Elasticsearch BM25 via Reciprocal Rank Fusion

LangGraphQdrantElasticsearchNeo4jRAGASFastAPI

◈ Click to see details

Key Featuresclick to flip back

Production RAG System

  • ▸Built hybrid retrieval combining Qdrant dense search and Elasticsearch BM25 via Reciprocal Rank Fusion
  • ▸Achieved 23.7% retrieval faithfulness gain with context precision of 1.0 across 50 evaluated queries at 163.5ms mean latency
  • ▸Implemented CRAG as a LangGraph state machine with query rewriting and web search fallback, triggering self-correction on 38% of queries
  • ▸Achieved 0.87 answer relevancy across the arXiv research corpus evaluation with Neo4j graph relationships for contextual depth
  • ▸Integrated RAGAS evaluation dashboard monitoring faithfulness, context precision, and answer relevancy in real time via FastAPI and React
LangGraph · Qdrant · Elasticsearch · Neo4j · RAGAS · FastAPI
GitHub ↗
Apr 2025 — PresentFeatured

LiveWire AI Meeting Transcription

Dual-channel audio capture (tab + mic) via Chrome MV3 extension and Python WebSocket server

Chrome MV3FastAPIWebSocketsWhisperasyncioFFmpeg

◈ Click to see details

Key Featuresclick to flip back

LiveWire AI Meeting Transcription

  • ▸Dual-channel audio capture (tab + mic) via Chrome MV3 extension and Python WebSocket server
  • ▸Async FFmpeg + asyncio for parallel transcription streams across Zoom, Teams, and Google Meet
  • ▸Platform fingerprinting, rejoin recovery, and RMS trend tracking for production resilience
  • ▸EBML magic byte validation for clean WebM audio headers and session registry for concurrent recording
  • ▸Production-deployed at VoiceBotics AI with device switch detection and health timeline logging
Chrome MV3 · FastAPI · WebSockets · Whisper · asyncio · FFmpeg
GitHub ↗
Jan 2026 — Feb 2026Featured

Job Command Center

Full-stack AI job search automation — scrapes LinkedIn & Wellfound via Apify with configurable filters

FastAPIReactClaude HaikuApifySQLiteLaTeX

◈ Click to see details

Key Featuresclick to flip back

Job Command Center

  • ▸Full-stack AI job search automation — scrapes LinkedIn & Wellfound via Apify with configurable filters
  • ▸Claude Haiku AI scores each job by profile fit on a 0–100 scale with reasoning explanation
  • ▸Auto-generates tailored LaTeX resumes and LinkedIn InMail messages per job description
  • ▸React dashboard with job pipeline, notes, application status tracking, and fit score sorting
  • ▸FastAPI + SQLite backend with Vercel-deployed frontend — handles full job search workflow end-to-end
FastAPI · React · Claude Haiku · Apify · SQLite · LaTeX
GitHub ↗
Jan 2023 — May 2023Research

Real-Time Vehicle Detection (CVR Journal)

Developed a YOLOv8 object detection model processing 5,000+ video frames with 88% accuracy for vehicle counting and classification

PythonYOLOv8OpenCVTensorFlowPyTorch

◈ Click to see details

Key Featuresclick to flip back

Real-Time Vehicle Detection (CVR Journal)

  • ▸Developed a YOLOv8 object detection model processing 5,000+ video frames with 88% accuracy for vehicle counting and classification
  • ▸Optimized inference by 25% using TensorFlow model quantization and OpenCV pipelines for real-time edge performance
  • ▸Enabled deployment on low-power hardware while maintaining high-speed frame analysis for traffic surveillance
  • ▸Published in CVR Journal of Science and Technology Vol. 24, June 2023 — secured 3rd Prize at Project Expo 2K23
  • ▸Complete pipeline: data collection → training → inference → real-time annotated video output
Python · YOLOv8 · OpenCV · TensorFlow · PyTorch
GitHub ↗
Jun 2022 — Dec 2022

Face Recognition Attendance System

Automated attendance system processing live camera feeds for real-time multi-face identification

PythonOpenCVface_recognitionSQLiteDeep Learning

◈ Click to see details

Key Featuresclick to flip back

Face Recognition Attendance System

  • ▸Automated attendance system processing live camera feeds for real-time multi-face identification
  • ▸Face embedding store for fast identity matching with configurable confidence thresholds at runtime
  • ▸Automated attendance logging with timestamps and session data directly to SQLite database
  • ▸Multi-face detection within a single frame — handles crowded classrooms and office environments
  • ▸Unknown person alerting system with confidence-based flagging for security and access control
Python · OpenCV · face_recognition · SQLite · Deep Learning
GitHub ↗
Jan 2024 — May 2024

AI-Powered Vehicle Detection & Tracking for Surveillance

PP-YOLO-based object detection and multi-object tracking for real-time traffic surveillance systems

PythonPP-YOLOOpenCVNumPyComputer Vision

◈ Click to see details

Key Featuresclick to flip back

AI-Powered Vehicle Detection & Tracking for Surveillance

  • ▸PP-YOLO-based object detection and multi-object tracking for real-time traffic surveillance systems
  • ▸Benchmarked 8% mAP improvement over baseline through improved NMS, anchor tuning, and augmentation
  • ▸Extended class support: cars, trucks, pedestrians, cyclists, and emergency vehicles
  • ▸Frame-by-frame trajectory estimation with lane-level tracking and directional flow analysis
  • ▸Supports both video file and live webcam input with annotated output and confidence overlays
Python · PP-YOLO · OpenCV · NumPy · Computer Vision
GitHub ↗
Jan 2024 — May 2024

Shamir's Secret Sharing

Cryptographic threshold scheme via Lagrange polynomial interpolation over Mersenne Prime finite field

PythonCryptographyFinite FieldsPolynomial Interpolation

◈ Click to see details

Key Featuresclick to flip back

Shamir's Secret Sharing

  • ▸Cryptographic threshold scheme via Lagrange polynomial interpolation over Mersenne Prime finite field
  • ▸Configurable k-of-n: split secret into N shares, reconstruct with any K subset — information-theoretically secure
  • ▸Galois Field GF(2^8) arithmetic ensures zero information leakage from any subset smaller than threshold
  • ▸CLI interface for share generation, secure distribution, and secret reconstruction workflows
  • ▸Comprehensive unit tests verifying correctness across boundary k and n values and edge cases
Python · Cryptography · Finite Fields · Polynomial Interpolation
GitHub ↗
Jan 2025 — May 2025

Hospital Management System

Full-stack MVC hospital platform: patient records, appointment scheduling, billing, and staff management

JavaMySQLMVC ArchitectureRBACSoftware Engineering

◈ Click to see details

Key Featuresclick to flip back

Hospital Management System

  • ▸Full-stack MVC hospital platform: patient records, appointment scheduling, billing, and staff management
  • ▸Role-based access control with scoped permissions for admin, doctor, and receptionist roles
  • ▸Conflict detection engine prevents double-booking with real-time doctor availability checking
  • ▸Automated invoice generation per visit with itemized billing breakdown and payment tracking
  • ▸Developed for Advanced Software Engineering — includes full documentation and reliability analysis
Java · MySQL · MVC Architecture · RBAC · Software Engineering
GitHub ↗
Jan 2024 — May 2024

Earthquake Impact Prediction

Cloud-based ML ensemble predicting structural damage severity from seismic and geospatial features

PythonScikit-learnAWSPandasCloud ML

◈ Click to see details

Key Featuresclick to flip back

Earthquake Impact Prediction

  • ▸Cloud-based ML ensemble predicting structural damage severity from seismic and geospatial features
  • ▸Feature engineering: magnitude, depth, soil type, fault distance, and historical damage weighting
  • ▸Ensemble model: Random Forest + Gradient Boosting with voting for multi-class damage classification
  • ▸Geospatial analysis with distance-to-fault proximity weighting for improved regional accuracy
  • ▸Full evaluation suite: ROC curves, confusion matrix, feature importance, and cross-validation
Python · Scikit-learn · AWS · Pandas · Cloud ML
GitHub ↗
Aug 2024 — Dec 2024

Machine Learning Coursework

ID3 Decision Trees, Logistic Regression, SVM, Lasso & Ridge Regression built from scratch in NumPy

PythonNumPyScikit-learnNLPJupyter Notebook

◈ Click to see details

Key Featuresclick to flip back

Machine Learning Coursework

  • ▸ID3 Decision Trees, Logistic Regression, SVM, Lasso & Ridge Regression built from scratch in NumPy
  • ▸Neural network with full backpropagation and gradient descent implemented without ML frameworks
  • ▸Sentiment analysis pipeline using NLP preprocessing, TF-IDF, and multiple classifier comparison
  • ▸Kernel PCA, LDA, and dimensionality reduction with explained variance and 2D projection plots
  • ▸K-Means, DBSCAN, and Hierarchical clustering with silhouette scoring and visual cluster output
Python · NumPy · Scikit-learn · NLP · Jupyter Notebook
GitHub ↗
Jan 2024 — May 2024

Distributed Cloud Computing

AWS-based distributed system with multi-node architecture, leader election, and automatic failover

AWSPythonDistributed SystemsCloud ArchitectureJupyter

◈ Click to see details

Key Featuresclick to flip back

Distributed Cloud Computing

  • ▸AWS-based distributed system with multi-node architecture, leader election, and automatic failover
  • ▸Cloud architecture implementation with fault tolerance and exponential backoff retry logic
  • ▸Round-robin and weighted load balancing strategies with horizontal scaling at runtime
  • ▸Performance benchmarking: throughput, latency, and scalability tradeoff analysis across node counts
  • ▸Distributed computing experiments comparing centralized vs decentralized coordination approaches
AWS · Python · Distributed Systems · Cloud Architecture · Jupyter
GitHub ↗
Jan 2025 — May 2025

APL Mini Language Interpreter

Custom lexer and recursive-descent parser generating a full AST from APL array programming syntax

PythonASTParsingCompiler DesignCS609

◈ Click to see details

Key Featuresclick to flip back

APL Mini Language Interpreter

  • ▸Custom lexer and recursive-descent parser generating a full AST from APL array programming syntax
  • ▸Array primitives: reshape, iota, reduce, scan, outer product, and transpose operations
  • ▸Monadic and dyadic function dispatch with operator precedence and associativity resolution
  • ▸Built for CS609 Advanced Programming Languages — demonstrates compiler design fundamentals
  • ▸REPL interface with session history, error reporting with line/column context, and type checking
Python · AST · Parsing · Compiler Design · CS609
GitHub ↗
Jan 2024 — May 2024

AI Maze Search Visualizer

Interactive visualizer for BFS, DFS, A*, and Dijkstra search algorithms with live animation

PythonPygameA* SearchBFS/DFSAI Algorithms

◈ Click to see details

Key Featuresclick to flip back

AI Maze Search Visualizer

  • ▸Interactive visualizer for BFS, DFS, A*, and Dijkstra search algorithms with live animation
  • ▸Procedural maze generation with configurable wall density, grid size, and start/end placement
  • ▸Step-by-step animation with adjustable speed — each node visit animated in real time
  • ▸Side-by-side algorithm comparison: path length, nodes explored, and time complexity metrics
  • ▸Educational tool demonstrating heuristic vs uninformed search with performance benchmarking
Python · Pygame · A* Search · BFS/DFS · AI Algorithms
GitHub ↗
Jan 2025 — May 2025

IIoT Research Survey

Systematic literature review of 50+ papers on Industrial IoT security, ML, and edge computing

ResearchIIoTEdge AIFederated LearningSecurity

◈ Click to see details

Key Featuresclick to flip back

IIoT Research Survey

  • ▸Systematic literature review of 50+ papers on Industrial IoT security, ML, and edge computing
  • ▸Analysis of ML-based anomaly detection approaches for industrial sensor streams and SCADA systems
  • ▸Edge vs cloud inference tradeoff evaluation for real-time ICS environments with latency constraints
  • ▸Taxonomy of attack vectors in industrial control systems with mitigation strategy classification
  • ▸Future directions analysis: federated learning for privacy-preserving IIoT deployments at scale
Research · IIoT · Edge AI · Federated Learning · Security
GitHub ↗
Aug 2024 — Dec 2024

HCI Case Studies

Usability testing sessions with task completion rate, error rate, and time-on-task metrics

UX DesignUsability TestingWCAG 2.1FigmaHCI

◈ Click to see details

Key Featuresclick to flip back

HCI Case Studies

  • ▸Usability testing sessions with task completion rate, error rate, and time-on-task metrics
  • ▸WCAG 2.1 accessibility audit with severity ratings and actionable remediation recommendations
  • ▸A/B interface experiments measuring cognitive load and user preference across design variants
  • ▸Heuristic evaluation using Nielsen's 10 usability principles with prioritized issue backlog
  • ▸Low and high-fidelity prototypes iterated based on think-aloud user feedback sessions
UX Design · Usability Testing · WCAG 2.1 · Figma · HCI
GitHub ↗
Jan 2025 — May 2025

Traffic Monitoring AI

Computer vision pipeline for real-time vehicle movement extraction from live traffic video feeds

PythonOpenCVComputer VisionNumPyJupyter

◈ Click to see details

Key Featuresclick to flip back

Traffic Monitoring AI

  • ▸Computer vision pipeline for real-time vehicle movement extraction from live traffic video feeds
  • ▸Frame-by-frame object tracking with trajectory estimation and lane-level directional flow analysis
  • ▸Vehicle counting per lane with density classification: free flow / congested / gridlock states
  • ▸Preprocessing pipeline with contrast normalization and frame stabilization for edge hardware
  • ▸Output: annotated video overlays with per-lane counts, speed estimates, and density heatmap
Python · OpenCV · Computer Vision · NumPy · Jupyter
GitHub ↗
Jan 2026 — Feb 2026

Portfolio Website

Built with Next.js 16 + TypeScript + Tailwind CSS — 18-card flip project showcase with category filter

Next.jsTypeScriptTailwind CSSReactVercel

◈ Click to see details

Key Featuresclick to flip back

Portfolio Website

  • ▸Built with Next.js 16 + TypeScript + Tailwind CSS — 18-card flip project showcase with category filter
  • ▸Live terminal emulator with whoami, skills, projects, contact commands fully functional in browser
  • ▸JD keyword matcher running 100% client-side — paste any job description for instant skill match score
  • ▸Typewriter animation, scroll-triggered counters, custom cursor with lag ring, and scroll progress bar
  • ▸Fully responsive with IntersectionObserver fade-ins — deployed on Vercel with automatic CI/CD
Next.js · TypeScript · Tailwind CSS · React · Vercel
GitHub ↗

02.Background

Apr 2025 — Present

AI Systems Developer Intern @ VoiceBotics AI

  • →Developed a FastAPI WebSocket server with 60s reconnect window, achieving 94% preflight success rate for persistent streaming across high-latency environments
  • →Built an STT latency watchdog with automated recovery, maintaining mean transcription TTF under 2.5s across live Zoom and Teams sessions
  • →Implemented 16 normalized error codes with JSONL instrumentation logging, generating automated beta metrics summaries
  • →Engineered the LiveWire Chrome extension using tabCapture API and FFmpeg stereo splitting for dual-channel audio capture
Jun 2022 — Dec 2022

Software Engineer Intern @ Globalshala

  • →Managed Java and SQL data systems for cloud-native Azure integrations, maintaining 99.9% uptime across enterprise data ingestion pipelines
  • →Built Azure Databricks ETL pipelines for Power BI dashboards, delivering real-time KPI insights and reducing report generation time by 70%
  • →Optimized SQL schemas across distributed enterprise systems, improving data retrieval speeds by 40%
Jan 2024 — Dec 2025

MS, Computer Science @ Southeast Missouri State University

3.9 GPA. Specialization in Advanced AI, Distributed Systems, and Human-Computer Interaction.

2019 — 2023

BE, Computer Science @ CVR College of Engineering

Engineering foundation in Algorithms, Data Structures, Database Systems, and OOP.

📄 Peer-Reviewed Publication

UniLLMOps: A Unified Framework for End-to-End LLM Production Systems

Zenodo · DOI: 10.5281/zenodo.19582347 · April 2026

Read on Zenodo →
📄 Peer-Reviewed Publication

Real-Time Video-Based Vehicle Detection, Counting & Classification System

CVR Journal of Science & Technology · Vol. 24 · June 2023 · 3rd Prize Expo2K23

Read on ResearchGate →

03.Capabilities

Python / SQL / C95%
LLMs / RAG / Agents92%
ML (TF / PyTorch / Scikit)90%
CV (OpenCV / YOLO)88%
FastAPI / WebSockets85%
Cloud (AWS / Azure)82%
Cryptography78%
React / Next.js76%

04.Live Data