Yashswi Shukla

FINAL-YEAR COMPUTER SCIENCE UNDERGRADUATE AT BENNETT UNIVERSITY

I build AI systems that work in the real world, not just in demos — spanning computer vision, natural language processing, and production pipelines. My work is backed by a published research paper and hands-on experience building inside a startup Founder's Office.

Featured AI Systems

Built to Solve Real Problems

These are the strongest projects in my portfolio where I designed the AI workflow, implemented the system, and tracked meaningful outcomes.

Computer Vision Safety System

Suraksha Dristi

Problem: In classroom and public spaces, silent distress signals often go unnoticed and emergency response can be delayed.

System Architecture

Video ingestion from webcam and IP camera streams.
Gesture and face detection with MediaPipe and OpenCV.
Distress trigger pipeline to alert administrators in real time.

Tech Stack: Python, OpenCV, MediaPipe, Haar Cascade, NumPy

Results: 85% gesture recognition accuracy at ~18 FPS and 40% faster simulated emergency response in controlled tests.

GitHub Demo On Request

NLP Summarization Pipeline

Narrative

Problem: Analysts and students struggle to process long social discussion threads quickly and extract useful insights.

System Architecture

Reddit data collection with PRAW and cleaning pipeline.
NLP preprocessing and topic extraction with spaCy.
Abstractive summarization using transformer-based BART models.

Tech Stack: Python, Transformers, spaCy, PRAW, Streamlit

Results: Produced concise structured summaries from long-form discussions, reducing manual analysis time for multi-post threads.

GitHub Demo On Request

Generative AI Product

AI Resume Builder

Problem: Job seekers often produce generic resumes that fail ATS screening and do not communicate technical depth.

System Architecture

Structured input collection with guided workflow UI.
LLM-assisted drafting and iterative content refinement.
Programmatic PDF generation with formatting and export controls.

Tech Stack: Python, Streamlit, OpenAI API, FPDF

Results: Delivered a production-ready tool that generates ATS-oriented resumes with customizable sections and robust PDF output handling.

GitHub Live Demo

About and Engineering Philosophy

I Build AI Systems That Can Be Used, Not Just Presented

I am most energized by applied AI problems where model quality, latency, and usability all matter at the same time. My approach is to break a problem into system components: data handling, model choice, evaluation, deployment constraints, and product integration.

I care about engineering tradeoffs. Instead of optimizing only for model novelty, I focus on whether a system is reliable in real scenarios and whether it creates measurable value for users.

Problem First

Start from a concrete user pain point and define success criteria before model selection.

Metrics Driven

Track system behavior with clear evaluation metrics and iterate based on data, not assumptions.

Build End to End

From prototype to deployable app, I work across modeling, backend logic, and product execution.

Technical Stack

Tools I Use to Build and Ship AI Products

Core AI and ML

PyTorch TensorFlow Keras Scikit-learn NumPy Pandas

Computer Vision

OpenCV MediaPipe Image Enhancement Face Detection Noise Reduction

NLP

Transformers BART spaCy Text Summarization Prompt Engineering

Backend and App Development

Python Streamlit Node.js Express.js JavaScript REST API Integration

Tools

Git GitHub Jupyter VS Code Postman EmailJS

Languages

Python Java C C++

Experience

CXO1.ai | Founder's Office, Automation & Growth Internship

Startup execution across automation, QA, growth, and client delivery for AI-visibility audits.

Founder's Office, Automation & Growth

Apr 2026 - Present

Automation Pipeline

Built and operate a nine-stage Python pipeline that automates AI-visibility audits by querying ChatGPT, Claude, and Gemini through their APIs, replacing a manual process capped at about 24 audits a day.
Wired in Anthropic, OpenAI, and Google Gemini APIs, enforcing a hard per-audit cost ceiling after usage-based costs proved harder to predict than expected.
Built the scoring engine behind each audit -- five weighted metrics (Citation Frequency, Position, Accuracy, Entity Clarity, Technical Accessibility) computed from 15 parsed query-platform responses per company.
Generated print-ready HTML templates for client-facing audit reports and follow-up action plans, auto-populated per engagement.

Methodology, QA & Validation

Helped implement Citation Architecture, a four-layer AI-visibility framework (Entity Foundation, Content Retrievability, Source Authority, Technical Accessibility) underlying every audit.
Validated the pipeline against live companies, with automated scores landing within a point of manual baselines on early runs.
Found and fixed a data-integrity bug that reported directory profiles as absent based on an indirect signal rather than a verified check, replacing it with a tiered confidence system now standard across every report.
Resolved further defects surfaced through live testing, including a JSON-parsing crash and a print-pagination bug in the report templates.

Go-to-Market & Operations

Researched target industries and helped define a seven-segment ICP framework spanning startups, SaaS, FinTech, legal, and healthtech companies.
Ran cold email outreach across four ICP segments using Instantly.ai to support B2B client acquisition.
Contributed to a nine-file internal knowledge base covering methodology, market intelligence, and delivery procedures for the practice.
Authored audit reports, competitive analyses, and internal documentation to support client delivery.

Skillcase | Founder's Office Internship

Startup execution across product ownership, system building, and day-to-day cross-functional delivery.

Product and Operations Intern

July 2025 - Feb 2026

Product Ownership

Owned requirement translation for mobile app and web platform versions.
Prioritized feature rollout across V1 to V3 with founder and developer alignment.
Validated deliverables through structured UAT and release readiness checks.

System Building

Supported end-to-end launch of a learning app used by active students and teachers.
Drove regression testing, bug reporting, and post-release issue tracking loops.
Helped shape mobile-first web updates for usability and reliability improvements.

Cross-Functional Execution

Acted as communication bridge between founders, developers, and operations teams.
Managed scheduling and student workflows for live class operations.
Created educational social content to support growth and brand engagement.

Certificate

Research

Enhancing Face Detection in Low-Light Conditions

Published work focused on denoising strategies for robust face detection in low-illumination scene at ICCSAI-2025 Conference.

Research Question

How do different denoising techniques affect face detection quality when images are degraded by low-light noise?

Method

Benchmarked Median, Non-Local Means, Wavelet, and DnCNN methods using PSNR, SSIM, and MSE on low-light datasets.

Key Finding

DnCNN delivered best restoration quality with PSNR up to 32.45, improving downstream face detection robustness.

Research Repository Publication PDF Certificate

Projects

Additional Systems and Prototypes

KholiPoll

Campus communication platform prototype with event feeds, feedback loops, and role-aware access.

Stack: Python, Streamlit, Data Processing

GitHub Live Demo

Story Visualizer

AI-assisted storytelling interface that combines narrative flow and visual presentation in an interactive app.

Stack: Python, Streamlit, PIL

Code On Request Live Demo

Currently Building

Active AI Engineering Work

In Progress

Low-Light Vision Evaluation Harness

Expanding robustness benchmarks with repeatable experiments across noise profiles and denoising configurations.

In Progress

NLP Summarization API Version

Converting the Narrative pipeline into an API-first service with cleaner modular interfaces and test coverage.

Exploring

AI Product Telemetry

Designing lightweight monitoring for prompt quality, user flow drop-off, and inference latency in AI apps.

Contact

Hiring for AI or ML Roles?

I am looking for full-time AI Engineer and Machine Learning Engineer opportunities where I can contribute across model development and system implementation.

yashswishukla@gmail.com
+91 8368528828
Gautam Buddha Nagar, Uttar Pradesh