I'm Vedant Singh. I work on Python backend systems, APIs, automation pipelines, and applied AI—grounded in verifiable source code, robust data workflows, and real-world utility.
Real code, not just claims. Each featured project is backed by verified GitHub source code, distinct architecture, and an authentic case study.
VIDEO REEL
A dual-API architecture decoupling conversational retrieval from transaction execution. Combines ChromaDB vector search over historical commodity records, local Ollama DeepSeek R1 reasoning with chain-of-thought stripping, and thread-safe purchase logging.
VIDEO REEL
End-to-end Python automation extracting contact leads from scanned document images using Tesseract OCR, enhancing email tone and markdown structure via OpenRouter (Mistral-7B), and dispatching through Microsoft Outlook COM with rate-limiting delays.
VIDEO REEL
A passenger assistance prototype developed at CRIS (Centre for Railway Information Systems) converting UTS ticket numbers, origin stations, and destinations into standardized Grade-1 Braille Unicode glyphs for visually impaired rail travelers.
VIDEO REEL
Interactive Streamlit dashboard exploring multidimensional socio-economic intelligence data. Implements 6 dynamic Plotly visual representations and an embedded Scikit-learn Random Forest Regressor predicting event likelihood with real-time parameter tuning.
Active web tools and microservices hosted on live production subdomains at sshconnect.space. Built with strict performance benchmarks, clean code architectures, and instant browser execution.
High-speed, browser-based document compiler and clean HTML parser. Seamlessly ingests Microsoft Word (.docx) binary archives, strips bloated Word XML styling, extracts semantic hierarchy, formats tables natively, and generates production-ready, sanitized HTML with real-time live preview and code export.
Documents are decompiled and converted strictly in volatile RAM memory streams. No client document payloads are ever committed to disk storage, guaranteeing confidentiality.
Curated repositories from @vineetvedant. Featured only where source code implementation is verifiable.
Custom-trained YOLO computer vision model trained on Roboflow MRI scans to detect and localize 5 distinct brain tumor categories (glioma, meningioma, pituitary, space-occupying lesion, no tumor). Includes evaluated confusion matrices and inference outputs.
Real-time webcam application combining OpenCV Haar Cascade facial detection with DeepFace convolutional neural network analysis to predict dominant emotional states and monitor on-screen vs. off-screen user focus time.
Lightweight offline Python desktop utility featuring a Tkinter GUI and PyPDF4 text extraction, allowing users to select PDF documents and listen via pyttsx3 offline text-to-speech with selectable voice gender and speech-rate controls.
Zero-dependency client-side HTML cleaner and document converter with live split-pane preview, whitespace stripping, and tag filtering options. Fully automated with Cloudflare Pages and Vercel deployment workflows.
This portfolio is curated for engineering substance rather than raw repository counts. Two repositories from the profile—object-detection-thermal---grayscale---RGB- (Teledyne FLIR experimental dataset documentation) and search-engine (UI mockups)—were audited but excluded from primary code showcases because functional source code was not committed to those repositories. Honest constraints prevent inflating prototype status into commercial claims.
From enterprise backend workflows to computing infrastructure and practical machine learning systems.
Python backend development, API integration, workflow automation, and engineering support in an enterprise technology environment.
Worked with HPC environments and performance-oriented engineering workflows, including computing infrastructure and Linux-based systems.
Developed computer vision systems, Python industrial automation, RAG workflows, and IoT monitoring applications that moved from prototypes into practical deployments.
Explored convolutional neural networks and thermal imaging for human detection in industrial contexts.
Built an accessibility-focused ticketing concept converting UTS ticket records into Braille representations for visually impaired passengers.
Continuous learning through industry-recognized certifications and academic publications in leading science venues.
Advanced concepts in cloud architecture, distributed computing, and scalable system design.
Comprehensive cybersecurity fundamentals including threat detection, security protocols, and risk management.
End-to-end data analytics including data collection, cleaning, analysis, and visualization techniques.
Highest level certificate in the NCC, demonstrating leadership, national security knowledge, physical endurance, and discipline.
Published in ICAIN 2025 · Springer LNNS (Lecture Notes in Networks and Systems) · Scopus Indexed
I'm Vedant, a software engineer based in Delhi. What interests me isn't collecting technology names—it's understanding systems well enough to make them simpler, more reliable, and useful for the people depending on them.
I've worked across backend APIs, machine learning, industrial devices, and high-performance computing. The common thread is engineering software that connects different pieces of a real-world problem.
Outside the code, serving as a Senior Under Officer in the National Cadet Corps taught me a lot about discipline, responsibility, and following through.
Not proficiency percentages—just technologies I've used in engineering work and verified public repositories.
Python · C++ · SQL · JavaScript · FastAPI · Uvicorn · REST APIs · SQLite
RAG · ChromaDB · Ollama (DeepSeek R1) · Mistral-7B · YOLOv8 · Scikit-learn · Pandas · NumPy · DeepFace
Linux · Git & GitHub Actions · Win32 COM · Tesseract OCR · Docker · Terraform · Raspberry Pi
Streamlit · Plotly Express · Tkinter · HTML5 / Vanilla CSS