CRM analytics
Working with postal CRM data to explore complaint patterns and support meaningful reporting.
Work includes validating complaint counts, investigating report behaviour, and improving the clarity of data presented to users.
Hasna C A
AI EngineerResearch Scholar
Building AI for real-world systems.
Exploring how machines see, reason, and solve.
Centre for Excellence in Postal TechnologyAI engineering
NIT CalicutPart-time PhD research
01 / The intersection
An engineer's approach.
A researcher's curiosity.
I’m Hasna C A, working at the intersection of applied artificial intelligence and multimodal visual reasoning. At CEPT, my work includes CRM analytics and AI reports. Alongside this, I’m pursuing a part-time PhD in Computer Science and Engineering at NIT Calicut.
Applied AICentre for Excellence in Postal Technology, Kochi
Visual reasoningPart-time PhD · CSED, NIT Calicut
Research guidanceDr. Chandramani Chaudhary
Public service perspectiveDepartment of Posts, India
02 / Selected work
From data to decisions.
Professional work and hands-on exploration across analytics, language models, and intelligent workflows.
AI reports
Exploring natural-language access to data and making AI-generated reports more reliable.
Areas of work include report summaries, CSV downloads, end-to-end testing, and request logging using Postman and SigNoz.
Tool-using AI agents
Connecting language models with tools to execute multi-step tasks and select relevant actions.
Implemented a LangGraph and MCP workflow with Llama 3.1 8B, exploring automatic tool selection and testing behaviour on specific use cases.
Medical language model adaptation
Hands-on exploration of parameter-efficient fine-tuning and retrieval-augmented generation.
Fine-tuned Qwen2.5-3B using QLoRA on a 10,000-example subset of the ChatDoctor HealthCareMagic dataset. This is a learning project; no clinical deployment is claimed.
03 / Research in progress
Seeing is only
the beginning.
How can AI move from recognising an image to reasoning about it?
My research interests centre on multimodal visual reasoning, visual logic puzzles, and chain-of-vision methods—how models can connect visual evidence with structured reasoning.
Part-time PhD · Computer Science & Engineering
National Institute of Technology Calicut
Interactive demonstrations
6 challengesVisual reasoning lab
Explore spatial transformations, pattern completion, and rule inference through six small challenges.
Reflection
Reflect the green pattern from left to right. Draw the result in the output grid.
Test input Observe
Your answer Apply the rule
Select cells to build your answer.
Study the examples, then build your answer.
Original demonstration puzzles inspired by visual reasoning tasks.
Research areas
Questions guiding my doctoral research.
Multimodal visual reasoning
How models can combine visual and textual evidence to reason about a task.
Visual logic puzzles
How models can infer patterns, rules, and relationships from visual demonstrations.
Vision-language models
How visual representations and language can be aligned to support reasoning beyond recognition.
Chain-of-vision methods
How intermediate visual steps can support multi-step problem solving and make reasoning easier to inspect.
04 / Academic foundation
A continuing pursuit.
PhD, Computer Science & Engineering
National Institute of Technology Calicut
Part-time Research Scholar
M.Tech, Data Science & Artificial Intelligence
Cochin University of Science and Technology
CUSAT
B.Tech, Computer Science & Engineering
Government Engineering College Idukki
Painavu, Kerala
05 / AI engineering skills
AI engineering, in practice.
From language models and visual reasoning to practical AI applications.
Generative AI & LLM applications
LLM applications for natural-language access to operational data and AI-generated reporting.
LLMsNLPPython
AI reportsRAG & knowledge systems
Retrieval-augmented generation that brings relevant source documents and context into language model responses.
RetrievalEmbeddingsContext grounding
Medical RAG learning projectAgentic AI & tool orchestration
Tool-using workflows with LangGraph, MCP, and Llama 3.1 8B, including automatic selection of relevant tools.
LangGraphMCPTool selection
Tool-using AI agentsLLM fine-tuning
Parameter-efficient adaptation of Qwen2.5-3B using QLoRA on a 10,000-example medical dataset.
LoRA / QLoRAQwen2.54-bit quantization
Language model adaptationComputer vision & multimodal AI
Doctoral research interests in multimodal visual reasoning, visual logic puzzles, and vision-language models.
Computer visionVision-language modelsVisual reasoning
Research areasDeep learning & transformers
Neural network and transformer techniques supporting language modelling and computer vision.
Neural networksTransformersModel evaluation
Visual reasoning researchAI application engineering
Python and FastAPI for practical AI workflows, supported by end-to-end testing and request log inspection.
PythonFastAPIREST APIs
AI reporting applicationsData analytics & NL-to-SQL
SQL querying and CRM complaint analytics, alongside natural-language-to-SQL reporting for postal data.
SQLCRM analyticsNL-to-SQL
CRM analytics06 / Start a conversation
Interesting problems.
Thoughtful collaboration.
Personal email
hasnaca1991@gmail.comFor conversations about applied AI, visual reasoning, and research collaboration.