GULLINKALA MOULI SAGAR

AI Engineer

Key Strengths

Strong foundational knowledge and practical experience in AI, ML, and Deep Learning, particularly in Computer Vision and Generative AI.
Demonstrated ability to build and deploy end-to-end ML pipelines, from data curation to edge inference.
Hands-on experience with various LLMs (Llama 3, LLaVA, GPT-40-mini) and RAG architectures (ChromaDB, FAISS).
Proficiency in prompt engineering, agentic design, and model optimization techniques (quantization, TensorRT).
Experience with cloud services (AWS S3, SageMaker) for scalable model training and deployment.
Proven ability to work on real-world applications through internships and academic projects, including hackathon achievements.

Cultural & Operational Fit

Cultural Fit Analysis

The candidate's diverse project portfolio, ranging from news crawlers and study companions to defect detection and mental health chatbots, demonstrates a broad interest in applying AI across various domains. The involvement in hackathons and academic projects, alongside internships, indicates a proactive and learning-oriented mindset. The target role of AI Engineer aligns well with the candidate's demonstrated technical depth in AI/ML, particularly in areas like LLMs, RAG, and Computer Vision. The breadth of skills, including programming, ML frameworks, cloud, and deployment, suggests adaptability to different project requirements.

Soft Skills & Operational Fit

The candidate's project descriptions indicate a structured approach to problem-solving and a focus on practical application. The experience in team-based hackathons suggests an ability to collaborate. The detailed descriptions of technical challenges and solutions imply good communication of technical concepts. However, without direct interview data, a comprehensive assessment of soft skills like leadership, conflict resolution, or adaptability is limited.

AI is analyzing your overall score…

Identifying your key strengths…

Evaluating your skill match against the job requirements…

Assessing your cultural and operational fit

About

A Graduate Engineer with a strong passion for AI, Machine Learning and Deep Learning - actively building and deploying real-world applications over the past 1.5 years through internships, academic projects and national hackathons. Experienced across the end-to-end ML lifecycle: data curation, feature engineering, CNN and Transformer architecture design, training pipelines (PyTorch, TensorFlow, Keras), hyperparameter tuning and deployment-ready inference. Hands-on with edge deployment on NVIDIA Jetson Nano and Raspberry Pi using TensorFlow Lite, ONNX and TensorRT. Working knowledge of AWS cloud services - S3 for dataset storage, SageMaker for managed model training and Lambda for serverless inference triggers. Actively building with large language models and generative AI - running Llama 3 and LLaVA locally via Ollama, designing RAG pipelines with ChromaDB and FAISS, and developing agentic AI workflows using LangChain, CrewAI and the OpenAI API. Comfortable with prompt engineering practices including system prompt design, context grounding strategies and parameter tuning (temperature, top-p) to control LLM behaviour for production use cases.

Top Skills

PythonTensorFlowKerasPyTorchOpencv

Projects

Real-Time Sign Language Translator

January 1, 2026 – January 1, 2026

• Built ASL alphabet recognition system using MobileNetV2 (TensorFlow/Keras) on ASL Alphabet dataset (87,000 images, 29 classes) - integrated MediaPipe hand landmark detection to crop hand ROI before classification, improving accuracy over full-frame inference. • Deployed dual inference modes: static image upload and live webcam stream (1.5-second frame capture) with sentence builder converting recognised letters to words in real time. • Applied transfer learning by freezing early MobileNetV2 layers and systematically fine-tuning top 15 layers with data augmentation and dropout regularisation to prevent overfitting and improve robustness on gesture recognition tasks across diverse hand positions and lighting conditions.

View Project

Mental Health Companion - AI Powered Emotional Support Chatbot

November 1, 2025 – November 1, 2025

• Built empathetic conversational AI with GPT-40-mini - designed system prompt with non-clinical supportive tone, session memory across 20 turns and low temperature for measured, consistent responses. • Implemented agentic response pipeline where messages flow through mood classification, crisis detection and context management layers before LLM generation - providing structured decision-making logic beyond a simple chatbot. • Engineered real-time crisis detection using keyword matching to flag self-harm or suicidal ideation, triggering immediate safety response with verified helpline resources - concurrent mood classification (anxious, sad, angry, positive) updates live UI indicator.

View Project

RAG-Based Private Study Companion

November 1, 2025 – December 1, 2025

• Built fully offline RAG pipeline with Llama 3 8B (Ollama) and ChromaDB - grounds answers strictly in user-uploaded materials (PDF, TXT, MD) with zero external API dependency. • Implemented recursive text chunking (400-token chunks, 40-token overlap) with nomic-embed-text semantic embeddings (768-dim vectors) and cosine similarity retrieval with 0.72 cutoff to filter low-confidence chunks. • Developed system prompt engineering approach with low temperature, explicit source-citation instructions and hard refusal directives to ground responses - tuned chunk size, overlap and retrieval count parameters based on observed correlation between retrieval precision and response factuality.

View Project

Regional NewsCrawler - Agentic RAG Pipeline for Live News Q&A

September 1, 2025 – June 1, 2026

• Developed end-to-end data curation pipeline: crawls articles (The Hindu, Indian Express, Times of India) via Playwright, deduplicates with SHA-256 and MinHash LSH (Jaccard ≥ 0.8), embeds into FAISS index, serves Q&A via Groq's Llama 3.3-70B. • Designed unified RAG prompt strategy with context boundaries and citation instructions - LLM uses retrieved chunks when relevant, otherwise uses general knowledge. Built resilient fallback chain that auto-selects available Groq models on deprecation or rate-limit. • Built and deployed Flask application with "Surprise Me" feature (LLM-generated questions), live dataset statistics and Parquet export - currently serves via local Groq API (LPU-based inference) with architecture designed for AWS integration combining ultra-fast inference with cloud robustness.

View Project

Intelligent RoadEye - Road Surface Defect Detection & Analysis

October 1, 2024 – December 1, 2024

• Built VGG16-based patch classification pipeline (TensorFlow/Keras) for crack and pothole detection - curated dataset from drone captures, open datasets and web images, annotated with LabelImg, augmented with rotation, zoom and brightness shifts to address class imbalance. • Implemented sliding window inference (224x224 patches, 50% overlap) with GPU-accelerated CUDA training using early stopping, learning rate scheduling and dropout regularisation for stable convergence. • Built Flask dashboard with defect bounding boxes, summary panels, defect distribution chart and severity meter - evaluated via classification report and confusion matrix, fine-tuned last 4 VGG16 layers to improve validation accuracy.

View Project

Certifications

Advanced Computer Vision with TensorFlow

DeepLearning.AI / Coursera

May 1, 2026 – Present

Machine Learning for Computer Vision

MathWorks / Coursera

May 1, 2026 – Present

PCAP: Python Input, Output, and String Handling

Logical Operations / Coursera

April 1, 2026 – Present

Exploratory Data Analysis for Machine Learning

IBM / Coursera

November 1, 2024 – Present

IBM Introduction to Machine Learning Specialization

IBM / Coursera

November 1, 2024 – Present

Key Strengths

Strong foundational knowledge and practical experience in AI, ML, and Deep Learning, particularly in Computer Vision and Generative AI.
Demonstrated ability to build and deploy end-to-end ML pipelines, from data curation to edge inference.
Hands-on experience with various LLMs (Llama 3, LLaVA, GPT-40-mini) and RAG architectures (ChromaDB, FAISS).
Proficiency in prompt engineering, agentic design, and model optimization techniques (quantization, TensorRT).
Experience with cloud services (AWS S3, SageMaker) for scalable model training and deployment.
Proven ability to work on real-world applications through internships and academic projects, including hackathon achievements.

Cultural & Operational Fit

Cultural Fit Analysis

Soft Skills & Operational Fit

GULLINKALA MOULI SAGAR

Key Strengths

Cultural & Operational Fit

About

Top Skills

Skills

Education

Experience

Projects

Certifications

Key Strengths

Cultural & Operational Fit