Available for AI / Research collaboration

Hi, I'mGoutham Lal S H

>

Building Intelligent Vision Systems & AI Solutions. Deep learning for computer vision, production detection systems, and LLM-powered agents — from doctoral research to deployed code.

View ProjectsContact Me
Alappuzha, Kerala, India
Goutham Lal S H, AI Engineer and PhD Researcher

6+

Years Exp

PhD

Researcher

12+

AI Projects

About Me

Engineering intelligence, grounded in research

Where production machine learning meets doctoral computer vision.

0+

Years of Experience

0+

Flagship AI Projects

0+

Models Worked With

0

Workshops Delivered

I am an AI Engineer at Vuelogix Technologies and a PhD researcher at VELS University, working at the intersection of computer vision, deep learning and large language models. My day-to-day is building highly parallelized machine-learning applications in the cloud — object detection and tracking systems, document AI and OCR pipelines, vision foundation models, multi-agent LLM systems, and the MLOps and model-optimisation work that gets them into production.

My doctoral research develops computer vision frameworks for skin cancer detection using deep learning, focusing on making dermoscopic classification accurate enough, and interpretable enough, to be clinically useful.

Alongside engineering and research, I serve as a resource person for university faculty development programmes and workshops on Natural Language Processing and Machine Learning, having trained faculty and students across three multi-day programmes.

Core Strengths

Hard-workingTeam PlayerSincerePunctualityDedicatedCommunication SkillsResponsibleTime ManagementMotivator & Leader

Languages

English
Malayalam
Hindi
Tamil

AI Engineer @ Vuelogix

Cloud-scale machine learning applications, detection systems and LLM deployment.

PhD in Computer Science

Ongoing at VELS University, Chennai — computer vision for medical imaging.

Skin Cancer Detection Research

Deep learning frameworks for dermoscopic lesion classification and segmentation.

Resource Person & Educator

NLP and AI/ML workshops for CSI, Mahaguru Institute and KMEA College.

Technical Stack

Skills & Technologies

The tools I reach for across research, model development and production deployment.

Detection, segmentation, tracking and vision backbones — from research to real-time inference.

Object Detection (YOLOv8–11, RT-DETR)95%
Segmentation (SAM2, Mask R-CNN)90%
Multi-Object Tracking (ByteTrack, BoT-SORT)88%
Vision Transformers (DINOv2, ViT, Swin)87%
OCR & Document AI90%
Image Restoration & Enhancement82%

Full Toolset — Computer Vision & Detection

YOLOv8/9/10/11RT-DETRRF-DETRD-FINEGrounding DINOFaster R-CNNCascade R-CNNSSDSAM2DINOv2EVACLIPSigLIPFlorence-2ConvNeXt V2SwinViTByteTrackDeepSORTOC-SORTBoT-SORTStrongSORTPatchCorePaDiMFastFlowRestormerNAFNetReal-ESRGANOpenCV
Career Path

Experience

Six years across research labs, product engineering and the classroom.

CurrentApril 2023 — Present

Research Fellow

VELS University

Chennai, Tamil Nadu

Doctoral research on computer vision frameworks for skin cancer detection using deep learning.

  • Research topic: Computer Vision Frameworks for Skin Cancer Detection Using Deep Learning.
  • Design and evaluate deep convolutional and transformer architectures for dermoscopic lesion classification.
  • Build reproducible training and evaluation pipelines for medical imaging datasets.
PyTorchOpenCVDeep LearningMedical ImagingPython
CurrentSeptember 2021 — Present

AI Engineer

Vuelogix Technologies Pvt. Ltd.

Noel Focus, Kakkanad, Kochi

Building highly parallelized deep-learning applications deployed in the cloud.

  • Build highly parallelized deep-learning applications deployed in the cloud.
  • Apply machine learning and statistical modeling techniques to develop and evaluate algorithms that improve performance, quality, data management and accuracy.
  • Ship computer vision detection systems, OCR/eKYC pipelines and LLM-backed conversational services to production.
  • Design, train and fine-tune deep learning and computer vision models, including vision-language models (VLMs), for detection, segmentation and multimodal understanding.
  • Build and deploy LLM-backed agents and conversational services, from prompting and retrieval-augmented generation to fine-tuning and production serving.
PythonPyTorchTensorFlowYOLOTriton ServerFastAPIDockerRayKafkaKubernetes
March 2019 — September 2021

Junior Software Developer

Calculus Edupoint

West of Thiruvampady Jn, Alappuzha

Full-stack application development and data-driven internal tooling.

  • Developed and maintained web applications across the full stack.
  • Built database-backed modules and internal tools supporting daily operations.
PythonPHPMySQLJavaScriptHTML/CSS
November 2020 — September 2021

Computer Programming Faculty

Unique Occassio Tech

Thiruvampady, Alappuzha

Taught programming fundamentals and applied computer science to students.

  • Delivered instruction in programming fundamentals, data structures and databases.
  • Mentored students through hands-on projects and practical assessments.
PythonJavaC++SQL
Selected Work

Projects

Enterprise-grade vision, language and multimodal systems — architected, benchmarked, optimised and deployed.

Detection & Tracking

Universal Vision Detection Benchmark

A unified framework to train, tune and compare every modern object detector — YOLOv8–11, RT-DETR, RF-DETR, D-FINE, Grounding DINO and the R-CNN/SSD family — on custom datasets through one interface. Automatic hyperparameter tuning, multi-GPU and mixed-precision training, MLflow tracking and one-click ONNX/TensorRT export, scored on mAP, FPS, latency, memory, FLOPs and energy.

Use case: Choosing the right detector for a deployment by comparing accuracy, speed and cost on your own data — not a public leaderboard.

YOLOv11RT-DETRRF-DETRD-FINEGrounding DINOTensorRTMLflow
Request CodeDiscuss
Vision Research

Foundation Vision Models Lab

A research harness that evaluates vision foundation models — DINOv2, EVA, SigLIP, CLIP, SAM2, Florence-2 and ConvNeXt V2 — across zero-shot classification, image retrieval, few-shot learning, feature extraction and fine-tuning, so their transfer behaviour can be compared apples-to-apples.

Use case: Deciding which pretrained backbone to build on before committing to a downstream task.

DINOv2CLIPSigLIPSAM2Florence-2ConvNeXt V2
Request CodeDiscuss
Detection & Tracking

Industrial Defect Inspection Platform

A production inspection system combining detection, segmentation, anomaly detection (PatchCore, PaDiM, FastFlow) and OCR across multi-camera lines, with TensorRT-optimised inference and factory-floor connectivity over PLC, MQTT and OPC-UA.

Use case: Automated visual quality control on manufacturing lines, flagging defects in real time.

YOLO11SAM2PatchCorePaDiMFastFlowTensorRTMQTTOPC-UA
Request CodeDiscuss
Document AI

Document AI Platform

A unified document-understanding platform that parses invoices, passports, receipts, bank statements, driving licences and medical records using Donut, LayoutLMv3, Nougat, Florence-2, TrOCR and PaddleOCR — OCR, layout analysis and key-field extraction behind one API.

Use case: Straight-through processing of mixed document types in finance, KYC and healthcare back-offices.

DonutLayoutLMv3NougatFlorence-2TrOCRPaddleOCR
Request CodeDiscuss
Document AI

Vision-Language Document Intelligence

OCR-free document understanding built on vision-language models — Qwen2.5-VL, InternVL, MiniCPM-V, LLaVA and Florence-2 — covering table understanding, chart reasoning, document QA and form extraction directly from the pixels, without a separate OCR stage.

Use case: Answering natural-language questions over complex documents, tables and charts.

Qwen2.5-VLInternVLMiniCPM-VLLaVAFlorence-2
Request CodeDiscuss
Detection & Tracking

Multi-Modal Surveillance AI

A single real-time pipeline chaining detection, multi-object tracking, re-identification, pose estimation, face recognition, fall detection, weapon detection, and crowd and behaviour analysis — YOLO → ByteTrack → Pose → ReID → alert engine → dashboard.

Use case: Intelligent CCTV for public safety, retail and workplace monitoring with automated alerting.

YOLO11ByteTrackPose EstimationReIDFace Recognition
Request CodeDiscuss
Detection & Tracking

Multi-Object Tracking Benchmark

A benchmarking suite comparing modern trackers — ByteTrack, DeepSORT, OC-SORT, BoT-SORT and StrongSORT — on MOTA, MOTP, IDF1, HOTA and FPS to quantify the accuracy-versus-speed trade-off on custom video.

Use case: Selecting a tracker for a surveillance or analytics pipeline based on measured HOTA and throughput.

ByteTrackDeepSORTOC-SORTBoT-SORTStrongSORT
Request CodeDiscuss
Vision Research

CNN vs Transformer Study

A controlled comparative study of CNNs (ResNet, ConvNeXt, EfficientNet, DenseNet) against vision transformers (ViT, Swin, DeiT, EVA, DINOv2) across transfer learning, low- and large-data regimes, robustness, explainability and inference cost.

Use case: Evidence-based architecture selection for a new vision problem under real data and latency constraints.

ConvNeXtEfficientNetViTSwinDINOv2
Request CodeDiscuss
Document AI

Production OCR Suite

An end-to-end OCR platform — text detection and recognition, key-information extraction, table and signature detection, barcode/QR reading and handwriting recognition — built on PaddleOCR, TrOCR, Donut, Florence-2 and LayoutLM.

Use case: Digitising forms, receipts and handwritten records at scale with structured output.

PaddleOCRTrOCRDonutFlorence-2LayoutLM
Request CodeDiscuss
Vision Research

Image Restoration Framework

Deblurring, denoising, super-resolution, inpainting, dehazing, low-light enhancement and colourisation in one framework, spanning GAN and diffusion approaches — ESRGAN, Real-ESRGAN, NAFNet, Restormer, MIRNet and DiffBIR.

Use case: Recovering usable imagery from degraded CCTV, scanned documents and low-light captures.

Real-ESRGANNAFNetRestormerMIRNetDiffBIR
Request CodeDiscuss
Agentic & RAG

Enterprise Agentic AI Platform

A multi-agent system with specialised Planner, Research, Coding, Vision, SQL, Web-Search, Report-Writer, Reviewer and Memory agents, orchestrated with LangGraph, CrewAI, AutoGen and MCP — with human-in-the-loop approval, long-term memory, tool calling and event streaming.

Use case: Automating complex knowledge work that needs planning, tool use and review across many steps.

LangGraphCrewAIAutoGenMCPOpenAI Agents SDK
Request CodeDiscuss
Agentic & RAG

Enterprise Hybrid RAG Platform

A retrieval-augmented generation platform ingesting PDFs, Office documents, images, tables, databases and APIs, with advanced retrieval — Graph RAG, hybrid search, context compression, multi-query and re-ranking — served by Llama 3, Qwen, Mistral or Gemma over Milvus, Weaviate, Pinecone or pgvector.

Use case: Grounded question-answering and search across large private knowledge bases.

Llama 3Graph RAGHybrid SearchMilvusWeaviatepgvector
Request CodeDiscuss
Multimodal & Video

Vision-Language Agent

A multimodal agent that takes images, video, audio, PDFs and web pages, then detects, reasons, searches, explains, generates reports and executes actions — combining Qwen2.5-VL and GPT-4 Vision with Grounding DINO and SAM2 for grounded perception and tool calling.

Use case: An assistant that can look at real-world media, reason about it and act — not just chat.

Qwen2.5-VLGPT-4 VisionGrounding DINOSAM2Tool Calling
Request CodeDiscuss
Medical AI

Medical Imaging AI Platform

A clinical imaging platform for detection, segmentation, diagnosis support, automated report generation and a medical chatbot, built on MONAI, MedSAM, UNETR and SwinUNETR with LLaVA-Med for multimodal reasoning — an applied extension of my doctoral work in medical image analysis.

Use case: Assisting radiologists and dermatologists with segmentation, triage and structured reporting.

MONAIMedSAMUNETRSwinUNETRLLaVA-Med
Request CodeDiscuss
Multimodal & Video

Autonomous Video Intelligence

An end-to-end video-understanding pipeline: scene understanding → detection → tracking → action recognition → captioning → summarisation → natural-language query, powered by YOLO11, RT-DETR, VideoMAE, InternVideo2, SAM2 and Qwen2.5-VL.

Use case: Turning hours of footage into searchable, queryable events and summaries.

YOLO11RT-DETRVideoMAEInternVideo2Qwen2.5-VLSAM2
Request CodeDiscuss
MLOps & Optimization

AI Model Optimization Framework

A framework to optimise any model for deployment — ONNX, TensorRT, OpenVINO, TVM and TorchScript export plus quantization, pruning and distillation — with automated benchmarking across GPU, CPU, Jetson and Edge TPU targets.

Use case: Squeezing models to hit latency and memory budgets on the target hardware before shipping.

TensorRTONNXOpenVINOTVMQuantizationDistillation
Request CodeDiscuss
MLOps & Optimization

AI Experiment & Benchmark Dashboard

A web platform to track and compare hundreds of experiments — MLflow and W&B for tracking, Prometheus and Grafana for monitoring, FastAPI + React + PostgreSQL for the app — with ROC, PR curves, confusion matrices, mAP, latency, FLOPs and memory visualised side by side.

Use case: Giving a team one place to see which model and config actually won, and why.

MLflowW&BGrafanaFastAPIReactPostgreSQL
Request CodeDiscuss
Doctoral Work

Research

Making deep learning accurate enough — and interpretable enough — for clinical dermatology.

Ph.D. Thesis2023 — Present

Computer Vision Frameworks for Skin Cancer Detection Using Deep Learning

VELS Institute of Science, Technology & Advanced Studies, Chennai, Tamil NaduDoctor of Philosophy in Computer Science

Skin cancer is among the most treatable cancers when caught early and among the deadliest when missed. This research develops computer vision frameworks that classify dermoscopic images of skin lesions using deep learning — targeting the accuracy required for clinical relevance while keeping model decisions interpretable enough for a dermatologist to trust and audit.

01

Lesion Segmentation

Isolating lesion boundaries from surrounding skin so downstream classification sees signal rather than background artefacts.

02

Deep Classification Architectures

Comparative evaluation of convolutional and transformer-based backbones for multi-class dermoscopic classification.

03

Data Imbalance & Augmentation

Handling severe class imbalance across malignant and benign categories, where rare classes carry the highest clinical cost.

04

Interpretability

Surfacing the visual evidence behind each prediction, so clinical users can verify what the model attended to.

Research Interests

Medical Image AnalysisDeep LearningComputer VisionImage SegmentationTransfer LearningExplainable AI

Publications

Resource Person

Workshops & Training

Invited to lead multi-day faculty development programmes and workshops on NLP and machine learning.

5 Days
Faculty Development Programme

Natural Language Processing

Mahaguru Institute of Technology

Department of Computer Science and Engineering

Thiruvananthapuram, Kerala

In association with CSI (Computer Society of India)

5 Days
Workshop

Artificial Intelligence & Machine Learning

KMEA College, Aluva

Department of Computer Science and Engineering

Ernakulam, Kerala

In association with Aesthetix Edu-Tech, Ernakulam

5 Days
Workshop

Natural Language Processing

KMEA College, Aluva

Department of Computer Science and Engineering

Ernakulam, Kerala

In association with Aesthetix Edu-Tech, Ernakulam

Academics

Education

From a B.Sc. in Alappuzha to doctoral research in Chennai.

2023 — OngoingOngoing

Ph.D. in Computer Science

VELS University

Chennai, Tamil Nadu

Research topic: Computer Vision Frameworks for Skin Cancer Detection Using Deep Learning

2019 — 2022

Master of Computer Applications (MCA)

IGNOU University

SH College, Thevara

2017 — 2018

Master in Web Development Engineering

Kerala Computer Saksharatha Mission

IGI Alappuzha

2014 — 2017

B.Sc. in Computer Science

Kerala University

UIT Alappuzha

Get In Touch

Contact

Open to AI engineering roles, research collaboration and speaking invitations.

Location

Shanmugha Madoum, Allissery Ward, Alappuzha, Kerala 688001

This opens your email app with the message pre-filled. Prefer direct? gouthamchandu7736@gmail.com