Durgesh Dutt Sinha
AIML Engineer & Full-Stack Developer
Pune, Maharashtra, India
Professional Summary
Results-driven AI/ML Engineer & Full-Stack Developer with practical experience building and deploying intelligent web applications, computer vision tools, and machine learning pipelines. Creator of open-source projects including real-time multimodal voice assistants, e-commerce customer segmentation systems, and deep learning image classifiers. Proficient in Python, TypeScript, React, Next.js, FastAPI, and PyTorch, with a track record of delivering clean architectures from concept to live production.
Technical Skills
Languages: Python, TypeScript, JavaScript, SQL, HTML5, CSS3, C/C++
AI & Machine Learning: PyTorch, Scikit-Learn, Computer Vision (MediaPipe, CNN), Whisper STT, Gemini API, Pandas, NumPy, K-Means, DBSCAN
Full-Stack & Backend: Next.js, React, FastAPI, Streamlit, WebSockets, Node.js, RESTful APIs, Tailwind CSS
Cloud, Database & Tools: PostgreSQL (Neon), Drizzle ORM, Docker, Cloudflare Pages, Git & GitHub, AWS Educate
Featured Projects
Tech Stack: FastAPI • Python • WebSockets • React • TypeScript • Whisper STT • Gemini 2.0 Flash • Edge-TTS
- Engineered a full-duplex voice assistant streaming 16kHz audio over bi-directional WebSockets between React and FastAPI backend.
- Integrated Voice Activity Detection (VAD) with Whisper STT and Gemini 2.0 Flash streaming API, paired with Edge-TTS for low-latency voice playback.
Tech Stack: Next.js • React • TypeScript • Drizzle ORM • PostgreSQL (Neon) • Tailwind CSS • pdf-parse
- Developed full-stack platform parsing candidate resumes (.pdf/.docx) to evaluate alignment against 16 canonical tech role benchmarks.
- Implemented ATS compatibility scoring, keyword density extraction, and structured recommendations with Drizzle ORM persistence.
Tech Stack: Python 3.11 • Scikit-Learn • Streamlit • Plotly 3D • Pandas & NumPy • Joblib
- Built an end-to-end unsupervised ML pipeline processing 4,000+ retail records using RFM (Recency, Frequency, Monetary) feature extraction.
- Applied K-Means (K=5) and DBSCAN anomaly detection with log-scaling; built an interactive Streamlit UI with 3D Plotly visual projections.
Tech Stack: PyTorch • Python • Streamlit • Computer Vision • NumPy
- Trained a custom multi-layer CNN in PyTorch on Fashion-MNIST dataset, achieving ~91.4% test accuracy with dropout regularization.
- Deployed an interactive Streamlit web app enabling users to upload clothing images and receive real-time classification predictions.
Experience & Fellowships
AI Program Fellow — UNLOX®
Remote- Constructed and tested containerized AI deployment pipelines; evaluated autonomous agent workflows and machine learning models under technical mentorship.
AI Cohort Member — Be10x
India- Built prompt-chaining architectures and full-stack AI concept prototypes, demonstrating practical AI productivity automations.
Web Development Intern — InAmigos Foundation (IAF)
Remote- Developed and maintained responsive web pages and components for non-profit social initiatives, ensuring mobile responsiveness and accessibility.
Education
Master of Computer Applications (AIML)
Jul 2025 – 2027 (Expected)Sri Balaji University, Pune
Specialization: Artificial Intelligence & Machine Learning
Bachelor of Computer Applications
Aug 2022 – Apr 2025Sri Balaji University, Pune
Focus: AI, System Architecture, Prompt Engineering
Certifications & Community
- AWS Educate Emerging Talent Community MemberAmazon Web Services (AWS)
- AWS Educate Machine Learning FoundationsAmazon Web Services (AWS)