Portfolio

📍 New Delhi, India | 📧 Email | 📞 +91-9264436795 | 🔗 LinkedIn

---

👋 About Me

I’m a data-driven professional with a Computer Science foundation and an MSc in Business Analytics from the University of Birmingham. My experience bridges business analytics and machine learning, from building predictive and anomaly-detection models in SQL, Python, and PySpark, to developing secure LLM and RAG-based systems in production. I’ve delivered data solutions across consulting, healthcare, and technology domains that improve decision quality, automate reporting, and enhance operational intelligence. Passionate about unifying analytics with engineering, I thrive in roles that demand analytical depth, technical precision, and collaboration to transform complex data into actionable insight and scalable impact.

💡 Key Skills

Excel	Power BI	Python	SQL	Machine Learning	PyTorch	OpenAI
LangChain	FAISS	LLMs	Tableau	Looker Studio	AWS Cloud	GitHub

---

🎓 Education

MSc in Business Analytics | University of Birmingham, UK | Sep 2023 – Dec 2024
Grade: Merit
Modules: Data Management Strategies and Technologies, Predictive Modelling, Supply Chain & Logistics

B.Tech in Computer Science & Engineering | Guru Gobind Singh Indraprastha University, India | Aug 2018 – Aug 2022
Grade: Distinction
Modules: Statistical Methods in Computing, Machine Learning, Algorithms & Data Structures

---

💼 Experience

AI/ML Engineering Analyst Intern · Victoria Solutions
United Kingdom · Jul 2025 – Aug 2025

• Built production XGBoost churn prediction microservice processing 15K+ daily predictions with 94.2% accuracy for a telecom client; deployed using FastAPI + Docker achieving 42 ms P95 latency.
• Engineered MLOps pipeline with automated CI/CD, reducing deployment time from 6 hours to 18 minutes; implemented data drift monitoring via KS-test (α = 0.05) ensuring model reliability.
• Automated analysis of 80K customer support tickets using spaCy NLP + LDA topic modeling, cutting manual review from 240 hours → 24 hours (90% efficiency gain) and surfacing 7 actionable churn drivers.
• Designed MongoDB query optimization strategy lowering average retrieval latency from 340 ms → 68 ms across 2M+ records through compound indexing and aggregation pipelines.
• Implemented secure API authentication (JWT + rate-limiting 100 req/min) to protect endpoints from unauthorized access and potential adversarial attacks.

Data Analyst Inter · Blackmont Consulting
Cambridge, England, United Kingdom · Jan 2025 – Apr 2025

• Led an 8-member analytics team to develop reporting frameworks and visual storytelling dashboards for a global education client.
• Coordinated data extraction, transformation, and visualization across Excel, Power BI, and Tableau.
• Produced functional specifications and visual requirement documents supporting client presentations and executive reviews.
• Diagnosed data inconsistencies and optimized dashboard performance, improving reporting accuracy by 25%.

Strategic Design Consultant · Turner & Townsend
Birmingham, England, United Kingdom · Jun 2024 – Jul 2024

• Conducted customer data analysis to identify process inefficiencies, supporting data‑driven strategies for business growth and operational improvements.
• Partnered with delivery managers and IT teams to resolve 90% of identified workflow issues within project timelines.
• Contributed to feasibility assessments and business case development for internal analytics platforms.

Software Trainee · Entrepreneurship Cell, IIT Kharagpur
Nov 2021 – Jan 2022 · 3 mos

• Automated log‑file processing by developing Python scripts to ingest, parse, and validate system logs, reducing manual interventions by 40% and achieving an 80% success rate in nightly updates.
• Optimized database maintenance through refactoring SQL stored procedures and indexing strategies, improving update performance by 30% and ensuring consistent data integrity across all tables.
• Standardized support workflows by drafting step‑by‑step maintenance procedures, building error‑handling routines in Python and SQL, and delivering team training—resulting in a 25% decrease in incident resolution time.

---

🚀 Highlighted Projects

Competitor Analysis Engine — AI-Powered Market Intelligence Platform

Overview

Developed a comprehensive AI-powered competitor analysis platform that automates market research workflows using multi-step LLM orchestration, live web search, and visual analytics. Transforms 3-5 days of manual competitor research into actionable insights within 5-10 minutes. Designed for product managers, founders, and strategists who need real-time competitive intelligence without consulting fees.

---

✨ Key Features

🔍 Multi-Step Competitor Discovery
Employs GPT-4o with research-first methodology to discover 15-24 competitors across 4 distinct categories (market leaders, niche players, emerging startups, open-source alternatives).

🌐 Live Web Research Engine
Performs real-time DuckDuckGo searches with intelligent fallback logic, collecting 30+ search results per analysis to ensure comprehensive, up-to-date market data.

🧠 AI Feature Extraction
Leverages GPT-4o natural language processing to automatically identify, categorize, and compare 20-50+ product features across all competitors—from core capabilities to integrations.

📊 Interactive Visual Analytics
Generates 4 dynamic Plotly charts (market positioning bars, feature coverage pie, competitor treemap, comparison heatmap) with hover interactions and zoom capabilities.

🗂️ Feature Comparison Matrix
Creates spreadsheet-style matrices with color-coded cells (✅/❌/⚡) showing feature support across competitors, enabling quick gap analysis.

💡 Strategic Insights Generator
Uses GPT-4o to synthesize key differentiators, actionable recommendations, and missing capabilities—delivering executive-level strategic guidance.

📥 Multi-Format Export
Generates professional Excel spreadsheets (multi-sheet workbooks) and PowerPoint presentations (auto-formatted slides with charts) for offline sharing and presentations.

🎨 Premium UI/UX Design
Implements Microsoft Fluent + Apple Typography hybrid with frosted-glass cards, gradient-blur backgrounds, and responsive mobile layout (@768px breakpoint).

---

🛠️ Tech Stack

Python · Streamlit · OpenAI GPT-4o · DuckDuckGo Search · Pandas · Plotly · OpenPyXL · Python-PPTX

---

Key Contributions & Outcomes

✅ Built a modular multi-agent pipeline using async Python with 4-step GPT-4o prompts inspired by research analyst workflows (Hunter → Categorizer → Analyst → Reporter).
✅ Automated end-to-end competitor research from web scraping to export, reducing analysis time by 95% and costs by 99%.
✅ Implemented intelligent search fallback logic improving data quality by 50%.
✅ Increased competitor discovery output by 70% through optimized search collection.
✅ Engineered LLM caching using Streamlit @st.cache_data, cutting API costs ~60%.
✅ Designed production-grade CSS architecture (968 lines, 18KB) with zero Streamlit conflicts.
✅ Achieved WCAG AAA contrast via system-ui typography + frosted-glass UI.
✅ Continuous deployment using Streamlit Cloud + GitHub.
✅ Exported structured formats (Excel + PowerPoint) for stakeholder-ready reports.
✅ Implemented secure API key management with session-only storage.

---

📂 Repository

GitHub Repository

---

🌐 Live Demo

Live Streamlit App

---

📈 Impact Metrics

Metric	Value
Time Savings	95% reduction (3–5 days → 5–10 min)
Cost Savings	99% reduction ($5K–$50K → $1–$2)
Competitors Discovered	15–24 per analysis
Features Tracked	20–50+ per analysis
API Cost Efficiency	~$0.01–0.05 per report
Mobile Responsive	100% (768px breakpoint)
Export Formats	Excel + PowerPoint
Deployment Uptime	99.9%

---

🎯 Use Cases

• Product Managers: Feature prioritization, competitive audits
• Founders: Market validation, differentiation strategy
• Sales Teams: Battlecards, objection handling
• Marketing Teams: Positioning, USP development
• Investors/VCs: Due diligence automation

---

🔐 Security & Best Practices

• Zero credential persistence
• Input validation & sanitization
• Graceful error handling
• DuckDuckGo rate-limiting protection
• Full Git version control
• Modular backend architecture

---

🚀 Future Enhancements

• User authentication, saved analyses
• Slack/Teams/Discord integrations
• Additional data sources (Crunchbase, G2, LinkedIn)
• Sentiment analysis on reviews
• Weekly automated monitoring alerts
• RESTful API for enterprise usage
• Team collaboration workspaces

---

💼 Business Value:
Democratizes competitive intelligence by eliminating consultant costs, ensuring fresh market data, and scaling to any industry—from SaaS to fintech to e-commerce.

---

AI Daily Digest — Multi-Agent Newsletter Generator

Overview:
Developed a fully automated AI-powered daily news digest system leveraging multi-agent orchestration to fetch, summarize, and deliver the latest AI and tech headlines as elegant HTML newsletters. Designed to mimic human research workflows — from gathering credible sources to crafting polished summaries.

✨ Key Features:

• 🕵️‍♂️ Multi-Source RSS Hunter: Queries 8 verified RSS feeds (OpenAI, Google, TechCrunch, etc.) to retrieve the top 3 recent articles from each.
• 📚 Resilient Web Scraper: Uses adaptive User-Agent rotation and error-handling to bypass 403 restrictions and extract article text safely.
• 🧠 AI Summarization: Employs GPT-4o-mini with LangChain prompt templates to synthesize concise, 5-point summaries in clean HTML.
• 💌 Automated Email Composer: Injects AI summaries into a mobile-friendly HTML template and sends via SendGrid API.
• ⚙️ Smart Agent Workflow: Modeled after four personas — Hunter, Librarian, Editor, and Postman — using lightweight LangGraph-style coordination.

🛠️ Tech Stack:

           

Key Contributions & Outcomes:

• Built a multi-agent pipeline using modular Python classes inspired by LangGraph’s node-based architecture.
• Automated data ingestion and summarization across 8+ high-quality AI/Tech feeds, producing 200+ daily summaries.
• Implemented resilient scraping logic (feedparser + BeautifulSoup + requests) to handle blocked sources gracefully.
• Deployed email automation via SendGrid, achieving reliable delivery and 100% mobile-responsive rendering.
• Designed with environmental variable–based secrets management, ensuring secure credential handling in CI/CD.

📂 Repository: tanishqsharma7918/AI-Daily-Digest

---

RAG-MCP Chatbot — Contextual AI Assistant for ML Queries

Overview:
Developed an interactive Streamlit-based AI chatbot using Retrieval-Augmented Generation (RAG) and LangGraph agents to answer machine learning and data engineering queries contextually and reliably.

Tech & Tools:

              

Key Contributions & Outcomes:

• Engineered a multi-agent RAG architecture using LangGraph orchestration and FAISS vector stores for dynamic document retrieval.
• Built a knowledge-grounded chatbot using OpenAI GPT-4 API, enabling natural conversation with live contextual lookups.
• Designed modular RAG pipelines supporting document chunking, semantic search, and context memory management.
• Deployed via Streamlit with real-time UI for multi-turn interactions and API response tracking.
• Delivered a scalable foundation for enterprise-grade retrieval systems and educational chat assistants.

📂 Repository: tanishqsharma7918/RAG-MCP-chatbot

---

Winter Rock Ski Line Analytics

Duration: Feb 2024 – May 2024

• Overview: Analysed historical ski sales to uncover growth trends, seasonality, and supplier profitability under uncertain demand.
• Tech & Tools:
  Python
  Advance Excel
• Key Contributions & Outcomes:
  • Developed a Python‑driven centred moving average pipeline on 2019–2022 sales data, revealing a steady upward trend and pinpointing November–December as peak months.
  • Built Excel models to calculate seasonal indices and apply single exponential smoothing (α = 0.5), producing six‑month forecasts with a 3.77% MAE.
  • Designed a decision‑tree and Monte Carlo simulation in Excel/Python, determining the USA supplier yields a £32,500 higher expected profit and informing Winter Rock’s supplier selection.

---

Developing a Comprehensive NHS Dashboards: A Combined Approach for Management and General Audiences

Duration: Jan 2024 – Jun 2024

• Overview: Built two interactive dashboards—Hospital Patient Care Activity for management and Mental Health in England for the public—by blending NHS and UK Government data sources.
• Tech & Tools:
  Tableau Public
  Power BI
  Advance Excel
• Key Contributions & Outcomes:
  • Data Integration & Preparation: Cleaned, formatted, and blended NHS datasets with UK Government using Excel,Power BI and Tableau Prep, ensuring high data quality and consistency.
  • Management Dashboard: Designed interactive bar charts, scatter plots, and box plots to track admissions by specialty, length of stay, and waiting times—empowering NHS directors with actionable performance insights.
  • Public Dashboard: Created age‑ and gender‑specific prevalence visualizations and highlight tables to communicate mental health trends to a broad audience, improving public awareness and advocacy.
  • Design Excellence: Applied visualization best practices (clarity, consistency, interactivity) with filters and tooltips, resulting in user‑friendly dashboards published to Tableau Public for easy sharing.

---

Marketing Analytics Customer Segmentation

Duration: Feb 2024 – May 2024

• Overview: Conducted a full marketing‑analytics workflow on an airline passenger satisfaction dataset—data analysis, descriptive statistics, econometric modelling, and predictive classification.
• Tech & Tools:
  Python
  Advance Excel
• Key Contributions & Outcomes:
  • Defined a comprehensive data dictionary and performed data cleaning in Excel, delivering descriptive visualizations and summary statistics.
  • Developed multivariate and interaction regression models in Python (R²≈0.38), interpreting coefficients, marginal effects, and ANOVA statistics.
  • Built a logistic regression classifier, achieving 77.6% accuracy, 77.6% sensitivity, 77.6% specificity, and AUC = 0.8393, complete with ROC‑curve visualization.

---

Qualtrics-Driven BI Tool to Optimise Analytics for Birmingham International Academy

Duration: May 2024 – Sep 2024

• Overview: Developed and deployed a Qualtrics-driven BI tool, integrating automated data pipelines to enhance visualisation and improve reporting accuracy by 25%, significantly reducing manual errors and processing time.
• Tech & Tools:
  Advance Excel
  Jisc
  Qualtrics
  Asana
• Key Contributions & Outcomes:
  • Collaborated in a 5‑member team to deliver three BI solutions—Excel dashboard, Jisc survey dashboard, and Qualtrics dashboard—each tailored to distinct stakeholder groups and use cases.
  • Automated end‑to‑end data ingestion and transformation pipelines, cutting manual processing time by 30% and boosting data refresh frequency from weekly to daily.
  • Designed dynamic KPI visualisations and adaptive survey routing, yielding a 25% improvement in report accuracy and a 20% increase in actionable feedback uptake by BIA management.
  • Facilitated stakeholder workshops and feedback sessions, ensuring the final dashboards surpassed client expectations in usability and strategic insight delivery.

---

The Evolution of Scala: A GitHub History Analysis

Duration: Feb 2024 – Mar 2024

• Overview: Analysed over 30,000 commits spanning a decade of Scala development by extracting and cleaning data from Git and GitHub repositories, uncovering the contributors who have shaped the language and identifying key community experts.
• Tech & Tools:
  Python
  Advanced Excel

• Key Contributions & Outcomes:
  • Extracted and cleaned data from Git and GitHub repositories, processing >30,000 commits.
  • Mapped contributor activity to reveal development trends and highlight key experts in the Scala community.
  • Created clear visualisations in pandas and matplotlib to communicate insights on commit volume, contributor growth, and expertise distribution.

---

📈 Extracurricular & Leadership

Event Coordinator · Birmingham MoRun Marathon
Birmingham, UK · Nov 2024

• Successfully coordinated logistics for the event, managing parking for 900+ vehicles and guiding 1,200+ participants through the racing tracks.
• Efficiently handled the refreshment centre, catering to 1,000+ attendees, ensuring smooth operations and a positive experience for all in a high‑pressure, time‑sensitive environment.

Head Marshal · Winter Warmer Runs Birmingham
Birmingham, UK · Feb 2024

• Efficiently managed parking for around 200 cars, facilitated medal and food distribution, motivated runners, and supervised volunteers to ensure smooth event operations.

---

📫 Let’s Connect

Feel free to reach out for collaborations, questions, or opportunities!

• 📧 Email: tanishq.career@gmail.com
• 🔗 LinkedIn: https://www.linkedin.com/in/tanishq-sharma-/
• 🌐 Website / Portfolio: https://tanishqsharma7918.github.io/Portfolio/