Skip to content
AK
Back to homepage

Work Archive

All projects

The full set of AI systems and prototypes, grouped by primary expertise area.

Production-oriented prototype

LLM and Agentic AI

Enterprise Document Intelligence

Problem

Contracts and invoices arrive as unstructured documents that teams must read manually before any downstream system can use them.

System

An AI document-processing system that ingests documents, applies OCR and parsing, chunks and embeds the text, and uses retrieval with LLM orchestration to produce entity extractions and structured reports. Results are exposed through a FastAPI service for application integration.

My contribution

Designed the document-processing workflow, built the ingestion and preprocessing pipeline, implemented retrieval and LLM orchestration, developed structured API outputs, and created evaluation and error-analysis workflows.

  1. Documents
  2. Parsing and OCR
  3. Chunking
  4. Embeddings
  5. Retrieval
  6. LLM processing
  7. Structured output
  8. API

Technical focus

  • Retrieval-augmented generation
  • OCR and text parsing
  • Entity extraction
  • Structured outputs
  • Workflow automation

Stack

  • Python
  • LangChain
  • Transformer models
  • FastAPI
  • REST APIs
Case study: Enterprise Document IntelligenceGitHub repository for Enterprise Document Intelligence
Academic project

Computer Vision

Explainable Deepfake and Image Forgery Detection

Problem

Manipulated images and deepfakes are difficult to judge from a prediction alone, so a detector must also show why it flagged an image.

System

An explainable computer-vision system that detects manipulated images using deep learning and vision transformers, then produces attention-based explanation maps to localise the suspected manipulation. The work spans the DEEPFORGE final-year project and the related ViT-Xplain research on explainable detection.

My contribution

Built the image preprocessing pipeline, trained and evaluated the vision models, implemented attention visualisation for forgery localisation, and ran the error analysis used to compare model variants.

  1. Image input
  2. Preprocessing
  3. Vision model
  4. Manipulation prediction
  5. Attention or explanation map
  6. Forensic output

Technical focus

  • Deepfake detection
  • Forgery localisation
  • Vision transformers
  • Attention visualisation
  • Explainable AI

Stack

  • PyTorch
  • OpenCV
  • NumPy
  • Python
Case study: Explainable Deepfake and Image Forgery Detection
Research prototype

Computer Vision

Autonomous Vehicle Perception System

Problem

Autonomous-driving perception requires a unified spatial understanding of the scene from several camera and sensor streams.

System

A multi-camera perception prototype that preprocesses synchronised sensor streams, extracts features, builds a bird's-eye-view representation with BEVFormer, and performs 3D object detection and tracking. Built as a research prototype for autonomous-driving perception, not a vehicle deployment.

My contribution

Built the sensor data pipeline and preprocessing stage, integrated the BEV detection model, and set up the evaluation workflow used to compare detection and tracking configurations.

  1. Camera and sensor data
  2. Preprocessing
  3. Feature extraction
  4. BEV representation
  5. 3D detection
  6. Tracking and output

Technical focus

  • Multi-camera processing
  • 3D object detection
  • Sensor fusion
  • Object tracking
  • Real-time inference

Stack

  • PyTorch
  • BEVFormer
  • OpenCV
  • Python
Case study: Autonomous Vehicle Perception SystemGitHub repository for Autonomous Vehicle Perception System
Published research implementation

Predictive Machine Learning

Predictive Maintenance and Fault Detection

Problem

Industrial equipment failures surface late, when sensor drift has already turned into downtime.

System

A predictive-learning system that validates and preprocesses sensor time series, engineers temporal features, and applies residual and CRNN-style models to produce calibrated early-fault risk scores that feed maintenance alerts. It implements the direction of my published work on early-warning industrial fault detection.

My contribution

Designed the time-series preprocessing and feature pipeline, implemented and calibrated the fault-detection models, and built the evaluation pipeline used to report early-warning performance.

  1. Sensor data
  2. Validation and preprocessing
  3. Feature pipeline
  4. Time-series model
  5. Risk score
  6. Alert or maintenance recommendation

Technical focus

  • Predictive maintenance
  • Fault detection
  • Model calibration
  • Imbalanced learning
  • Anomaly detection

Stack

  • PyTorch
  • Scikit-learn
  • Pandas
  • NumPy
Case study: Predictive Maintenance and Fault DetectionPublication related to Predictive Maintenance and Fault Detection
Client project

LLM and Agentic AI

Multilingual Learning Assistant

Problem

Lecture recordings are hard to study from when learners do not share the language of the recording.

System

A multilingual learning workflow that extracts audio from lecture videos, transcribes it with speech recognition, translates and indexes the transcript, and then answers learner questions and generates quizzes from the indexed content. Built during my TLS work on learning-content automation.

My contribution

Built the transcript and translation pipeline, implemented the retrieval and question-answering workflow, and automated quiz generation from processed lecture content.

  1. Lecture video
  2. Audio extraction
  3. Speech recognition
  4. Transcript processing
  5. Translation and indexing
  6. Q&A and quiz generation

Technical focus

  • Speech recognition
  • Translation and multilingual NLP
  • Retrieval and question answering
  • Quiz generation
  • Workflow automation

Stack

  • Python
  • Transformer models
  • LangChain
  • FastAPI
Case study: Multilingual Learning Assistant
Production-oriented prototype

Predictive Machine Learning

Fraud and Anomaly Detection

Problem

Suspicious transactions are rare and shifting, so rule-based monitoring misses new fraud patterns.

System

A transaction-monitoring workflow that preprocesses transaction data, engineers behavioural features, and combines supervised ensemble models with anomaly detection under strong class imbalance. Scoring is exposed through an inference API for integration into application workflows.

My contribution

Built the transaction preprocessing and feature-engineering pipeline, trained and evaluated the ensemble and anomaly-detection models, and implemented the inference API.

  1. Transaction data
  2. Preprocessing
  3. Feature engineering
  4. Ensemble and anomaly models
  5. Risk score
  6. Inference API

Technical focus

  • Feature engineering
  • Ensemble learning
  • Anomaly detection
  • Imbalanced data handling
  • Inference APIs

Stack

  • Scikit-learn
  • XGBoost
  • Pandas
  • FastAPI
Case study: Fraud and Anomaly Detection
Research prototype

Computer Vision

Medical Imaging AI - Cancer Detection

Problem

Early signs of disease in CT and MRI scans are easy to miss during manual review.

System

A radiology-assistant prototype that preprocesses CT and MRI volumes and applies ResNet-based classification to highlight regions for radiologist review.

My contribution

Built the imaging preprocessing pipeline and trained and evaluated the classification models.

  1. Scan input
  2. Preprocessing
  3. CNN classification
  4. Region highlighting

Technical focus

  • Medical imaging
  • Image classification
  • Transfer learning

Stack

  • PyTorch
  • OpenCV
  • NumPy
Case study: Medical Imaging AI - Cancer Detection
Research prototype

Computer Vision

Retail Analytics & Customer Behavior

Problem

In-store conversion decisions rely on guesswork without visibility into how shoppers move and interact.

System

A computer-vision workflow that tracks customer journeys and product interactions using pose estimation, then aggregates the signals into behavioural analytics.

My contribution

Implemented the tracking and pose-estimation pipeline and the analytics aggregation layer.

  1. Camera feed
  2. Detection and pose estimation
  3. Tracking
  4. Analytics aggregation

Technical focus

  • Pose estimation
  • Object tracking
  • Behavioural analytics

Stack

  • PyTorch
  • OpenCV
  • Pandas
Case study: Retail Analytics & Customer BehaviorGitHub repository for Retail Analytics & Customer Behavior
Production-oriented prototype

Computer Vision

Manufacturing Quality Control Vision

Problem

Surface defects on a production line are hard to catch consistently by eye.

System

A defect-detection workflow using YOLOv8 on production-line imagery, designed for real-time inference at line speed.

My contribution

Prepared the defect dataset, trained the detector, and tuned inference for latency.

  1. Line camera
  2. Preprocessing
  3. YOLOv8 detection
  4. Defect alert

Technical focus

  • Object detection
  • Real-time inference
  • Model optimization

Stack

  • PyTorch
  • YOLOv8
  • OpenCV
Case study: Manufacturing Quality Control Vision
Research prototype

LLM and Agentic AI

Legal Document Analysis Platform

Problem

Contract review requires reading long documents to locate a small number of risk clauses.

System

An NLP workflow for contract review that segments documents, extracts named entities and clauses, and surfaces risk-relevant passages for a reviewer.

My contribution

Built the parsing and clause-extraction pipeline and the reviewer-facing output format.

  1. Contract input
  2. Segmentation
  3. Entity and clause extraction
  4. Risk summary

Technical focus

  • Named entity recognition
  • Information extraction
  • Document intelligence

Stack

  • Python
  • Transformer models
  • spaCy
Case study: Legal Document Analysis Platform
Research prototype

Research Projects

Agricultural Crop Health Monitoring

Problem

Crop disease and stress are detected too late when fields are inspected on the ground.

System

A segmentation workflow over drone and satellite imagery for early disease detection and crop-health mapping.

My contribution

Built the imagery preprocessing pipeline and trained the segmentation models.

  1. Aerial imagery
  2. Preprocessing
  3. Segmentation model
  4. Health map

Technical focus

  • Image segmentation
  • Satellite imagery
  • Remote sensing data

Stack

  • PyTorch
  • OpenCV
  • NumPy
Case study: Agricultural Crop Health Monitoring
Research prototype

LLM and Agentic AI

Content Moderation AI

Problem

Harmful content appears across text, image, and video at the same time.

System

A multimodal moderation workflow that scores text and imagery with CLIP-based representations and routes uncertain cases for human review.

My contribution

Implemented the multimodal scoring pipeline and the human-in-the-loop routing logic.

  1. Content input
  2. Multimodal encoding
  3. Classification
  4. Review routing

Technical focus

  • Multimodal AI
  • Content safety
  • Human-in-the-loop workflows

Stack

  • PyTorch
  • CLIP
  • Python
Case study: Content Moderation AI
Research prototype

LLM and Agentic AI

Sales Intelligence from Calls

Problem

Insights from sales calls stay locked in recordings nobody re-listens to.

System

A speech-to-insight workflow that transcribes calls with Whisper and uses an LLM to extract sentiment, topics, and action items.

My contribution

Built the transcription and post-processing pipeline and the structured extraction prompts.

  1. Call audio
  2. Speech recognition
  3. LLM analysis
  4. Structured summary

Technical focus

  • Speech-to-text
  • Sentiment analysis
  • Structured outputs

Stack

  • Whisper
  • Python
  • LangChain
Case study: Sales Intelligence from Calls
Research prototype

Data Engineering and MLOps

Smart City Traffic Management

Problem

Traffic incidents are noticed slowly when camera feeds are monitored manually.

System

A streaming pipeline for traffic-flow analysis and incident detection, with object tracking running at the edge and aggregated metrics processed downstream.

My contribution

Built the real-time data-processing pipeline and the edge inference integration.

  1. Camera streams
  2. Edge inference
  3. Stream processing
  4. Incident alerts

Technical focus

  • Edge computing
  • Object tracking
  • Real-time data processing

Stack

  • Python
  • OpenCV
  • Docker
Case study: Smart City Traffic Management