Durgesh Gowda C durgeshgowdac

Hey, I'm Durgesh 👋

I'm an AI/ML Engineer based in Bengaluru. I recently completed my MTech in Machine Intelligence at PES University (CGPA: 9.41, Prof. CNR Rao Scholar), and I build production AI systems — the kind that run in mission-critical environments, not just notebooks.

Most recently, I interned at the Indian Navy's AI Incubation Centre (INICAI), where I built a real-time multi-camera person re-identification system using NVIDIA DeepStream. Before that, I spent six months at ISRO (ISTRAC) engineering an NLP pipeline that gave scientists 10x faster access to satellite telemetry data.

What drives me

I'm drawn to the gap between AI research and real-world deployment. There's a version of machine learning that lives in papers and a very different version that has to survive noisy sensor feeds, mission-critical uptime requirements, and engineers who need answers in milliseconds. That gap is where I want to work.

At ISRO, I saw how domain experts couldn't access their own data quickly enough — I fixed it with NLP. At INICAI, I tackled one of computer vision's hardest problems: tracking a person as they disappear from one camera and reappear in another. These experiences shaped how I think about building AI: reliability and real impact first, elegance second.

Right now I'm exploring the GenAI stack — multimodal RAG, LLMs, and MLOps — and looking for AI Engineer roles where I can bring that production mindset to scale.

My tech toolbox

Here's the tech I use to bring ideas to life:

Languages — Python, C++, Java, Bash, C#
Machine Learning & Deep Learning — Machine Learning, Deep Learning, PyTorch, TensorFlow, Hugging Face Transformers, scikit-learn, TF-IDF
Computer Vision — OpenCV, YOLO, ByteTrack, FAISS, NVIDIA DeepStream, CUDA, Person Re-Identification, Object Detection & Tracking
NLP — Natural Language Processing (NLP), Named Entity Recognition (RoBERTa fine-tuning), Semantic Search, spaCy, NLTK, OpenAI Whisper
GenAI & Multimodal — Retrieval-Augmented Generation (RAG), Multimodal AI, Chainlit, pgvector, Gemini API, Vertex AI, Google Cloud AI, BigQuery ML
Data Science — Data Science, Data Analysis, Data Visualization, Pandas, NumPy, Matplotlib, Seaborn
Backend & Databases — FastAPI, Django, Next.js, Spring Framework, PostgreSQL, MySQL, SQLite, REST APIs, MVC
CS Fundamentals — Algorithms, Data Structures, Object-Oriented Programming (OOP), Multithreading, Regular Expressions, Debugging
Infrastructure & Cloud — Docker, Modal, Ollama, GCP, AWS
Tools — Git, GitHub, Postman, Linux

Currently building GenAI/cloud skills on Google Cloud (Vertex AI, Gemini API, BigQuery ML) — Diamond League, 84,700 pts.

Experience

1. Indian Navy — INICAI (AI/ML Engineer Intern)

Duration: October 2025 - December 2025 (3 months)

Real-Time Multi-Camera Person Tracking and Re-Identification

Architected and deployed a production-grade real-time multi-camera person Re-ID system using NVIDIA DeepStream SDK (Python + C++), integrating object detection, multi-object tracking, and appearance-based Re-ID into a single low-latency GStreamer inference.
Designed a multi-stage identity resolution pipeline maintaining consistent global IDs across camera handoffs, occlusions, and re-entries with robust distance aggregation.

2. ISRO — ISTRAC (AI/ML Project Intern)

Duration: February 2023 - July 2023 (6 months)

Satellite Data Retrieval, Processing & Visualization using NLP and AI

Fine-tuned RoBERTa for domain-specific NER on satellite mission vocabulary; integrated OpenAI Whisper for voice command support — giving scientists a natural language interface to ISRO's operational servers.
Built a semantic search engine (Whoosh + SpaCy) that replaced keyword-only lookup and improved retrieval speed 10x. Delivered as a production Tkinter desktop application inside ISRO's live environment.

3. Simple Icons (Open Source Contributor) View contributions →

Duration: Jan 2024 - present

Contributed several brand icons and updates to a widely used open-source icon library. Maintained JSON schema compliance via standard GitHub PR workflows. Continue to contribute occasionally.

Projects I'm proud of

Check out some of my work:

GlyphScholar — Multimodal RAG for PDFs. Parses text, tables, formulas, and page images (via MinerU), retrieves with dense vectors (pgvector), BM25, and image search in parallel, then reranks before answering through a Chainlit chat UI. Runs on Ollama locally or bursts to Modal GPUs. Live demo
FoundYou — Multi-camera person Re-ID with FAISS-accelerated three-stage identity matching, YOLO + ByteTrack + OSNet-512. Handles occlusions, re-entries, and cross-camera ID linking with minimal false positives. Co-authored research paper (Zenodo); MOTA 0.93 and IDF1 0.89 on a synthetic two-camera benchmark.
MLXnarrator — Offline narration pipeline for screen recordings on Apple Silicon. Speaks a plain-text script with Kokoro TTS (MLX), fits the audio to the video, generates subtitles with whisper.cpp, and renders the final MP4 with ffmpeg. Runs from the terminal or double-click launchers, with no cloud or API keys after a one-time model download.
RoleSense — Resume screening app that classifies resumes into 43 categories and matches them to job descriptions via TF-IDF similarity and a trained classifier (Random Forest, ~82% accuracy). Streamlit UI with batch processing and pass/fail thresholds. Live demo
ArtAttack — First-person 3D art gallery game in Unity with custom ShaderLab paint shaders, physics-based brush mechanics, security AI, and localization support.
SAMS — Full-stack Django attendance management system with role-based access (admin, teacher, student), dashboards, and reporting across institutions.
Newton-Gravity — Real-time 3D gravitational physics simulation in C++ and OpenGL with custom physics logic and interactive camera controls.
AVL Tree Visualizer — Interactive Java/Swing GUI for step-by-step AVL tree operations with controllable animation speed, built as an educational tool.

(Explore the repos for code and details!)

Hackathons & Competitions

Event	Project	Role	Result
CIDECODE Cyber Hackathon — Karnataka Police CID × PES University (2025)	NoFakeAI (deepfake detection system)	Team Lead	6th / 341 teams Top 40 nationally
TexPedition Hackathon — Epsilon x Hack2Skill (2025)	AdComply (Advertisement Compliance system)	Team Lead	Finalist
Smart India Hackathon — Ministry of Education (2025)	FixMyNagara (Civic Issue Reporting and Resolution System)	Team Lead	Selected
Tata Elxsi TELIPORT Season 3	Shipbox (Prompt-to-Package sandbox that turns prompts into a tested, dockerized zip)	Mentor	Mentored Team "Code Bunnies"

Education

M.Tech - PES University (2024 - 2026)

Computer Science Engineering (Machine Intelligence - Machine Learning + Artificial Intelligence)
Prof. CNR Rao Scholar
CGPA (10 points) : 9.41 (89.39 %)
GPA (4 points) : 3.76
UK Equivalent : First-Class Honours (1st) / Distinction
Participated in CIDECODE, TeXpedition, Smart India Hackathon (SIH), mentored in Tata Elxsi Hackathon (season 3)
Attended Guidewire DEVSummit 2025, and worked as Invigilation Teaching Assistant

B.Tech - Presidency University (2019 - 2023)

Computer Engineering (Data Science)
CGPA (10 points): 9.19 (87.30 %)
GPA (4 points) : 3.68
UK Equivalent : First-Class Honours (1st) / Distinction
Was part of Google Developer Student Club (GDSC), Microsoft Learn Student Ambassadors (MLSA), Science Club, and CodeChef Community.

Let's connect

Always happy to talk about AI, LLMs, RAG, AI agents, cool side projects, or the mysterious bugs that somehow disappear the moment you start debugging them.

LinkedIn: durgeshgowdac
Github: You're here! Star a repo if something catches your eye.

Little Inspiration

“First solve the problem. Then optimize it.” — Ken Thompson

Thanks for stopping by! 🚀

Miscellaneous [Click]

Profile Summary

🎓 MTech Computer Science (Machine Intelligence) — PES University · CGPA 9.41 · Prof. CNR Rao Scholar
🎓 BTech Computer Engineering (Data Science) — Presidency University · CGPA 9.19
🛰️ AI/ML Intern — ISRO ISTRAC & Indian Navy INICAI
☁️ Google Skills Diamond League — 84,700 pts · Gemini, Multimodal RAG, MLOps on Vertex AI, Terraform, Kubernetes
🏆 National hackathon finalist — 6th out of 341 teams (CIDECODE × Karnataka Police CID) · Selected for SIH 2025
💡 Mentored Team "Code Bunnies" at Tata Elxsi TELIPORT Season 3
💻 600+ LeetCode problems · 6,000+ Hyperskill problems · Open-source contributor to Simple Icons

Google Cloud — Skills & Certifications

Generative AI & LLMs

Develop Gen AI Apps with Gemini and Streamlit
Build Real World AI Applications with Gemini and Imagen
Inspect Rich Documents with Gemini Multimodality and Multimodal RAG
Explore Generative AI in Agent Platform · Prompt Design in Agent Platform
Introduction to Generative AI · Introduction to Large Language Models
Conversational AI on Vertex AI and Dialogflow CX
Responsible AI: Applying AI Principles with Google Cloud

MLOps & Machine Learning

Machine Learning Operations (MLOps) with Vertex AI: Model Evaluation
Certification Study Guide: Professional Machine Learning Engineer (PMLE)
Perform Predictive Data Analysis in BigQuery · Analyze Sentiment with Natural Language API
Use Machine Learning APIs on Google Cloud · Engineer Data for Predictive Modeling with BigQuery ML

Cloud Infrastructure & DevOps