Ajsal Shereef Palattuparambil
Ph.D. Candidate in Artificial Intelligence & Machine Learning Engineer
I design, build, and deploy intelligent systems that connect Reinforcement Learning, Generative AI, and Computer Vision—grounded in rigorous mathematics and focused on real-world impact.
Ajsal Shereef
Ph.D. Candidate · AI & ML
Reinforcement Learning
Adaptive agents, policy optimisation, and user-aligned behaviours.
Generative AI
LLM tooling, RAG, and QLORA-tuned models for real workflows.
Computer Vision on Edge
Segmentation, glare correction, and efficient models for constrained devices.
About
Bridging theory, models, and deployment.
I bridge deep theoretical expertise in Reinforcement Learning with practical applications in Generative AI and Computer Vision. I possess a strong command of the full AI lifecycle—from building RAG pipelines and Autonomous Agents to fine-tuning LLMs using QLORA. My work is underpinned by a rigorous mathematical foundation in probability, statistics, linear algebra, and optimization.
Focus Areas
- • Reinforcement Learning for adaptive, user-aligned decision-making.
- • Generative AI systems with RAG pipelines and tool-using agents.
- • Computer Vision models tailored for medical and edge settings.
- • Robust evaluation, benchmarking, and mathematical grounding.
Technical Skills
From foundational maths to production systems.
A toolkit that spans algorithm design, large-scale learning systems, and practical deployment on real hardware.
Core Languages
Generative AI & LLMs
Frameworks: LangChain, Hugging Face, Gradio
Computer Vision & Deep Learning
- • Object Detection
- • Segmentation (Yolact++, SoloV2, UNet)
- • GANs
- • VAEs
- • Diffusion Models
Reinforcement Learning
- • Policy Optimisation
- • Knowledge Transfer
- • Human-Agent Interaction
Deployment
- • On-device Deployment (Edge AI)
- • Quantization
- • Model Performance Benchmarking
Tools & Foundations
- • Git
- • LaTeX
- • Linear Algebra
- • Probability & Statistics
Projects
Implementing adaptive and transferable RL systems.
Selected research codebases that operationalise ideas in dynamic policy fusion and imagination-based knowledge transfer.
LLM Engineering
Hands-on engineering of LLM-powered systems, covering RAG pipelines, tool-using agents, and safe deployment patterns from prototype to production.
ASPECT: Analogical Semantic Policy Execution via Language Conditioned Transfer
A language-conditioned transfer framework that uses analogical semantics to execute policies in new tasks, extending ideas in knowledge transfer for RL with natural-language grounding.
Dynamic Policy Fusion for User Alignment Without Re-Interaction
A framework for personalising a pre-trained reinforcement learning policy to align with a user’s intent without re-interaction, using a theoretically grounded dynamic policy fusion approach.
Experience
Applying research to real-world systems.
A track record of building intelligent systems in industry and research labs, from medical imaging to call centre analytics.
- 2020 – 2022
Machine Learning Engineer
Trenser Technology Solutions
Engineered and deployed a real-time glare correction algorithm onto resource-constrained edge devices. Implemented and benchmarked SOTA computer vision models for medical image segmentation.
Edge CV • Medical Imaging • Benchmarking
- 2019 – 2020
Junior Data Scientist
Lotus Interworks
Developed mathematical models to evaluate call centre agent performance and contributed to an intelligent marketplace matching algorithm.
Predictive Analytics • Marketplace Modelling
- 2018 – 2019
Project Associate
IIT Madras
Modelled wicket probability for ESPN Cricinfo's smart stats project. Performed data analysis and statistical inference for environmental monitoring.
Sports Analytics • Environmental Monitoring
Teaching & Engagement
Mentoring and collaborative learning.
Experience supporting teaching delivery, QA sessions, and hands-on student engagement across university and community settings.
Graduate Research Teaching Fellow
Deakin University
Conducted workshop sessions and marked assessments.
Sessional Academic
Deakin University
Assisted the lecturer and handled QA sessions for course units.
Casual Research Assistant
Deakin University
Managed robotics lab equipment and demonstrated research prototypes to visiting school students.
Maths Tutor
EzyMaths
Private mathematics home tutoring for grade 9–12.
Education
Deep foundations in mathematics and learning.
Formal training that connects rigorous theory with practical machine learning systems.
Deakin University
Ph.D. in Reinforcement Learning
2022 – 2026
Thesis: Towards adapting Reinforcement Learning Agents.
IISER Trivandrum
BS-MS, Mathematics (Major) & Physics (Minor)
2013 – 2018
Thesis: AKS primality testing and cryptography.
Certifications
Continuous learning across ML and data science.
Formal certifications spanning data science fundamentals, TensorFlow for deep learning, and applied machine learning with Python and R.
CutShort Certified Data Science - Basic
CutShort · Issued Mar 2020
2019 AWS SageMaker, AI and Machine Learning - With Python
Udemy · Issued Nov 2019
ADVANCED COURSE IN ML USING PYTHON
GITAA · Issued Sep 2019
Sequences, Time Series and Prediction
Coursera · Issued Aug 2019
TensorFlow in Practice Specialization
Coursera · Issued Aug 2019
Convolutional Neural Networks in TensorFlow
Coursera · Issued Jul 2019
Introduction to TensorFlow for Artificial Intelligence, Machine Learning, and Deep Learning
Coursera · Issued Jul 2019
Natural Language Processing in TensorFlow
Coursera · Issued Jul 2019
Data Science for Beginners using R
GITAA · Issued Feb 2019
Research & Publications
Towards adaptive and aligned RL agents.
Selected work spanning knowledge transfer, dynamic policy fusion, and language-conditioned reinforcement learning.
MAGIK: Mapping to analogous goals via imagination-enabled knowledge transfer (ECAI 2025)
Personalisation via dynamic policy fusion (Intl. Conf. on Human-Agent Interaction 2024)
ASPECT: Analogical Semantic Policy Execution via Language Conditioned Transfer (ICML 2026 - Submitted)
Dynamic Policy Fusion for User Alignment Without Re-Interaction (Journal of Ambient Intelligence - Under Review)
Contact
Let's talk about intelligent systems.
Whether it's adaptive RL agents, generative AI workflows, or computer vision on edge devices, I'm happy to discuss collaborations, research, or applied projects.
Quick message