MSc Artificial Intelligence — UWE Bristol

VISHESH
JAKHAR

Vishesh Jakhar — AI Engineer, ML Engineer and Data Scientist based in Bristol, UK

I don't just build models — I build solutions that ship. From disease detection in agriculture to AI-powered real estate analytics, I turn research into production systems that solve real-world problems.

92%
Model Accuracy
75%
Size Optimization
4
Shipped Projects
Scroll

THE STORY

Every engineer has a starting point. Mine was a simple question: "What if AI could help farmers save their crops?"

I started where most AI engineers start — with curiosity. But curiosity without execution is just daydreaming. So I built things. Real things that real people use.

My first serious project wasn't for a grade — it was for farmers. A potato disease classification system that could identify crop diseases with 92% accuracy, optimized to run on mobile devices in areas with limited connectivity. That's when I realized: the best AI isn't the most complex — it's the most useful.

"I don't chase state-of-the-art benchmarks. I chase real-world impact. Every model I build answers one question: who does this help?"

From there, I went on to build production ML systems for real estate price prediction deployed on AWS, multi-agent AI systems that generate working code, and GenAI-powered tools for automated outreach. Each project taught me the same lesson — the gap between a notebook and production is where the real engineering happens.

Now, with an MSc in Artificial Intelligence from UWE Bristol, I bring together deep academic foundations in AI ethics, bias detection, and experimental research with hardcore engineering skills in Docker, cloud infrastructure, and scalable API design. I'm not just an ML engineer — I'm a full-stack AI problem solver.

EDUCATION

MSC ARTIFICIAL INTELLIGENCE
University of the West of England (UWE), Bristol
September 2024 — September 2025 · Classification: 2:1

Rigorous academic training blending theoretical ML research with hands-on system building. Covered advanced topics in deep learning architectures, AI ethics & governance, bias detection methodologies, hypothesis testing, and experimental design across 10+ research projects.

Advanced AIDeep LearningAI Ethics & GovernanceBias DetectionResearch MethodsStatistical AnalysisComputer VisionNLPReinforcement LearningData Engineering
92%
Highest Model Accuracy
75%
Model Size Reduction
0.85
R² Score Achieved
80%
Code Compilation Rate

TECHNICAL ARSENAL

Every tool here has been battle-tested in production — not just tutorials.

Languages & Core
PythonJavaScriptSQLHTML / CSSBash
Machine Learning & Deep Learning
TensorFlowPyTorchTF LiteKerasScikit-LearnOpenCVCNNsTransformersDiffusion ModelsData AugmentationModel Optimization
Generative & Agentic AI
LangChainLangGraphCrewAIGroqRAG PipelinesMulti-Agent SystemsPrompt EngineeringChromaDBVector Databases
Backend & APIs
FastAPIFlaskStreamlitREST APIsNode.jsReact
Cloud, DevOps & Infrastructure
DockerAWS EC2GCP Cloud FunctionsNginxCI/CDNetlifyLinuxGit / GitHub
Data Science & Analytics
PandasNumPyMatplotlibSeabornFeature EngineeringStatistical AnalysisEDA
AI Ethics & Research
Bias DetectionFairness MetricsExplainability (XAI)Responsible AIResearch Design

Proficiency Radar

SHIPPED PROJECTS

Each project tells a story: a problem I found, a solution I engineered, and the impact it created. Click to see them live.

01
Computer Vision · Production ML

POTATO DISEASE CLASSIFICATION

Farmers lose billions annually to undetected crop diseases. I built a deep learning system that identifies 3 potato leaf disease categories with 92% accuracy — optimized to run on mobile devices for field use in low-connectivity areas.

Problem

Potato diseases cause billions in global losses. Rural farmers lack access to plant pathologists for early diagnosis.

Solution

CNN model compressed 75% via TF Lite quantization for mobile edge deployment. Sub-200ms inference, works offline.

92%
Accuracy
75%
Model Compression
<200ms
Inference Latency
TensorFlowTF LiteFastAPIDockerGCP Cloud FunctionsReactData Augmentation
02
Regression Analysis · Full-Stack ML

BANGALORE HOME PRICES ML

Real estate pricing in Bangalore is chaotic. I built a machine learning system analyzing 13,000+ listings that predicts property prices with high accuracy — deployed on AWS with a clean web interface for real-time predictions.

Problem

Home buyers get exploited by opaque pricing. No accessible tool exists to estimate fair market value from real data.

Solution

Regression model with aggressive feature engineering (+23% boost). 2,500+ outliers removed. Deployed on AWS EC2 + Nginx.

0.85
R² Score
13K+
Listings Analyzed
+23%
Feature Eng. Impact
Scikit-LearnPandasNumPyFlaskAWS EC2NginxHTML / CSS / JS
03
Agentic AI · Multi-Agent Systems

CODER BUDDY

What if AI didn't just write code — but planned, architected, and debugged it like a team of engineers? A multi-agent system where specialized AI agents collaborate to generate complete, working applications from natural language.

Problem

Single-prompt code generation produces fragile, unstructured code. No planning, no architecture, no iteration.

Solution

3-agent pipeline: Planner → Architect → Coder. Each agent specializes, reviews, and iterates. 80% compilation on first pass.

80%
Compilation Rate
100+
Tokens / sec
3
AI Agents
LangGraphCrewAIGroqPythonMulti-Agent Orchestration
04
Generative AI · RAG Pipeline

COLD MAIL GENERATOR

Job seekers send generic emails that get ignored. This GenAI tool scrapes job postings, understands requirements via LLM, matches them to a portfolio in a vector database, and generates hyper-personalized cold outreach emails.

Problem

Generic cold emails have under 2% response rates. Personalization at scale is time-consuming and inconsistent.

Solution

RAG pipeline: scrape job pages → LLM extracts requirements → vector DB matches portfolio → generates tailored outreach.

RAG
Architecture
Auto
Job Scraping
Vector
Portfolio Match
LangChainChromaDBGroqStreamlitWeb ScrapingPrompt Engineering

THE JOURNEY

SEP 2024
STARTED MSC AI AT UWE BRISTOL
Began deep academic training in advanced AI, ethics, deep learning, and research methodology.
2024 – 2025
BUILT & SHIPPED PRODUCTION ML SYSTEMS
Deployed Potato Disease Classification (Netlify + GCP) and Bangalore Home Prices (AWS EC2) — real users, real traffic.
2025
ENTERED AGENTIC AI & GENERATIVE AI
Built Coder Buddy (multi-agent code generation) and Cold Mail Generator (RAG pipeline). Pushed into LangChain, CrewAI, vector databases.
OCT 2025
CODILITY SILVER AWARD — PI CHALLENGE
Earned Silver in Codility's competitive programming challenge, validating algorithmic problem-solving skills.
JAN 2026
NVIDIA DLI CERTIFIED × 2
Completed NVIDIA's Generative AI with Diffusion Models and Fundamentals of Deep Learning certifications.
NOW
WHAT'S NEXT?
Ready to bring this full-stack AI expertise to a team solving hard, meaningful problems. Your team, maybe?

CREDENTIALS

Industry-verified certifications backing every claim on this page.