AI/ML Developer - Computer Vision & LLMs

Harsha Priya Putsala

+91-939******* ******************@*****.*** linkedin.com/in/harshapriyaputsala github.com/HarshaPriya03

EDUCATION

B.Tech – Artificial Intelligence & Data Science (VIIT) 2020 – 2024

•Vignan’s Institute of Information Technology CGPA: 8.34

PROFESSIONAL EXPERIENCE

AI - ML Developer Anika Sterilis Pvt Ltd June 2024 – Present

Visakhapatnam, Andhra Pradesh

Real-Time Multi-Object Tracking & Crowd Analytics Python, YOLOv8, OpenCV, BoT-SORT, OSNet ReID

•Built a real-time multi-object tracking system using YOLOv8 and BoT-SORT with OSNet ReID to reliably track people through occlusions and crowded scenes without losing their identity.

•Designed a modular, config-driven architecture (config.yaml) that lets users easily tune confidence thresholds, tracker settings, and inference parameters without touching the code.

•Added unique people counting, frame-level JSON analytics logging, annotated video output, and command-line controls for flexible deployment.

•Evaluated YOLOv8n and YOLOv8m to find the right balance between tracking accuracy and processing speed.

Industrial CNC Machine Monitoring System Python, OpenCV, YOLOv8, PyTorch, ResNet18, FFmpeg, RTSP

•Built a real-time monitoring system for CNC machines using multiple computer vision models to track coolant sprinkler status, operator presence, and machine bed movement simultaneously.

•Trained a ResNet18-based classifier to detect coolant sprinkler ON/OFF states directly from video frames, making the system more robust than traditional motion-based detection.

•Used YOLOv8-based human detection to identify operator presence near CNC machines, supporting workplace safety monitoring.

•Developed a bed movement detection algorithm using segmentation centroid tracking between consecutive frames, cutting down false positives while accurately catching real movement.

•Integrated RTSP video streaming, FFmpeg encoding, and GPU inference, along with automated logging and real-time visualization for continuous monitoring.

Intelligent PAN & Aadhaar Information Extraction Python, EasyOCR, Gemini 2.5 Vision, OpenCV

•Built a document processing pipeline that automatically extracts structured data from PAN and Aadhaar cards.

•Combined EasyOCR with Gemini 2.5 Vision to boost extraction accuracy on blurry, rotated, or low-quality document scans.

•Extracted key fields including Name, Date of Birth, Gender, PAN Number, Aadhaar Number, Address, and Father's Name into clean, structured JSON output.

•Added post-processing and validation logic to catch and correct common OCR extraction errors.

AI-Powered Leave Management Chatbot Python, Flask, PHP, MySQL, Flan-T5, Sentence Transformers

•Engineered an NLP-based leave classification system using Flan-T5-Large achieving 95% accuracy in real-time reason analysis with customized ambiguity detection.

•Architected an auto-approval engine with multi-tier validation (monthly quota, balance prioritization, emergency flags), reducing HR workload by 80%.

•Architected a Flask + PHP hybrid backend with auto-failover supporting 150+ employees.

•Deployed semantic pattern detection via Sentence Transformers to identify leave abuse, achieving a 15% reduction in unauthorized leaves.

Policy Management System Python, Flask, MySQL, NLP, RAG

•Delivered a multi-agent HR chatbot pipeline with a multi-stage query resolution pipeline. Employee queries are first screened against active escalation rules (keyword/category-based) using bge-large-en, Qwen 2.5 14B instruct, then analyzed by Qwen 2.5 14B for multi-intent detection, it escalates when the query has complex queries that span multiple topics.

•Single-intent queries are matched against HR-created templates which are dynamically filled with employee details and returned instantly. If no template matches, ChromaDB performs semantic search over uploaded policy PDFs using Nomic embeddings, with LLaMA 3.1 8B Instruct generating answers from the retrieved chunks.

•Unresolved queries are automatically escalated to HR as pending tickets.

Power Management System Python, Flask, Plotly, websockets

•Engineered a real-time Power consumption monitoring system to track and optimize energy usage across facilities in real-time.

•Designed interactive dashboards tracking energy trends across 6 facilities for consumption trend visualization, driving data-driven decisions

Lead Scoring & CRM Query Assistant DistilRoBERTa, Flan-T5, Llama 3.1, AWS SageMaker

•Lead categorization system (Super Hot / Hot / Warm / Cool / Inactive) using DistilRoBERTa & Flan-T5, deployed on AWS SageMaker.

•Automated churn prediction using sentiment and activity signals, helping sales teams reduce lead drop-off by 50% to automatically classify whether the lead is getting churned or not based on negative_sentiment_count, inactivity_score and interaction count and it will be updated on every Sunday. It makes it easier for the Sales executive to concentrate more on the leads who are getting churned.

•AI-powered CRM query assistant using Llama 3.1 8B Instruct for accurate natural language to data retrieval from the tables regarding CRM.

PCB Defect Detection OpenCV, YOLO, Flask

•Engineered a real-time computer vision pipeline using YOLO + OpenCV, achieving 99% defect detection accuracy and reducing manual inspection time by several hours.

•Deployed a Flask-based web interface enabling real-time defect detection for production teams.

HR Management Chatbot Python, Flask, MySQL, NLP, Sentence Transformers, LLAMA 3.1 8B Instruct

•Architected a conversational AI agent for an HR Management System that answers natural language questions about employee leaves and leave balance tables. The system uses Llama 3.1 8B Instruct to generate MySQL queries from user questions, a SentenceTransformer-based intent detector to classify queries whether the query comes under leaves or leavebalance table.

•A custom date resolver to convert relative date phrases ("last month", "this week") into SQL-ready date ranges, status filtering (pending/approved/rejected), can also ask whether he is eligible to apply for leave or not, period-accurate leave day, automatic SQL validation and self-repair on errors, session-based pagination for large result sets, and LLM-powered verbalization of query results into natural language responses.

Resume extraction Python, Flask, NLP, Sentence Transformers, PdfPlumber, OpenAI

•Automated resume shortlisting using TF-IDF and cosine similarity, reducing manual screening time by 75%.

• Integrated auto-fill functionality that reduced manual data entry by 20% based on the uploaded resume data. Integrated with Flask to provide a web-based interface for real-time using openai and extraction tools .

•Based on the question asked by the user, it will give the output by analyzing the provided DB (using Langchain, openai).

Biomarkers Detection Python, Flask, NLP, Sentence Transformers, PdfPlumber, OpenAI

• Extraction of biomarkers, results and required range and tells the user if there is any excess or deficiency of the biomarker used pdfplumber and other extraction libraries to extract the content

Machine Learning Intern Spotmies – AU Incubation Centre Dec 2023 – Apr 2024

Visakhapatnam, Andhra Pradesh

•Developed an intelligent job description generator using FuzzyWuzzy token-set ratio for out-of-dataset job roles.

•Engineered a Live Streaming Avatar using the Wav2Lip pretrained model for real-time lip-sync from user text input.

•Deployed an HR conversational agent leveraging OpenAI APIs to analyze conversations and assess logical behavior.

Front-End Developer Intern East Coast Railway Jul 2023 – Sep 2023

Bhubaneswar Orissa

•Developed a Complaint Management System platform with admin resolution workflow, auto-escalation to dealers, and real-time status tracking for customers

•Using HTML, CSS, Bootstrap

Machine Learning Intern Coincent – Microsoft Jul 2022 – Sep 2022

Visakhapatnam, Andhra Pradesh

•Gained hands-on experience with NumPy, Pandas, Matplotlib, Scikit-Learn for ML model development.

•Delivered a diabetes prediction model achieving 83.4% accuracy, integrated with Django for real-time predictions.

RESEARCH & ACHIEVEMENTS

•3rd Prize in Web App Development – AgriTech category competition.

•I earned Azure- 900 Badge from Microsoft

•Participated in numerous coding competitions.

SKILLS

•Languages: Python, SQL, HTML, CSS

•ML / DL Frameworks: TensorFlow, PyTorch, Scikit-Learn, Hugging Face Transformers, NLTK

•Computer Vision: YOLOv8, OpenCV, Object Detection, Multi-Object Tracking (BoT-SORT), Re-Identification (OSNet), OCR, EasyOCR, Image Processing, RTSP Streaming

•NLP & Generative AI: Flan-T5, DistilRoBERTa, Sentence Transformers, Llama 3.1 8B Instruct, OpenAI GPT, LangChain, LangSmith, LangGraph, Qwen 2.5, Gemini 2.5 Vision, RAG, Prompt Engineering

•Data & Analytics: Pandas, NumPy, Matplotlib, Seaborn, pdfplumber

•Web & Backend: Flask, Django, Bootstrap, Figma, REST APIs

•Databases: MySQL

•Cloud & DevOps: AWS SageMaker, Git, GitHub, Docker

•Areas of Interest: Artificial Intelligence, Machine Learning, Natural Language Processing, Computer Vision, Generative AI

CERTIFICATIONS AND COURSEWORK

•Career Essentials in Generative AI – Microsoft

•Python with Data Science – Coincent, Microsoft (Jul 2022 – Sep 2022)

•IBM Skills Build – Project Management Fundamentals

•Database Management System – NPTEL (Jun 2022 – Aug 2022)

Contact this candidate