TALOSYN IT — DHAKA, BANGLADESH

We call it AI
only when it's actually AI.

We build applied AI systems for real operations — starting with healthcare — and we're fine-tuning open-weight models into a language model built specifically for Bangla. One standard across both: if it isn't AI, we don't call it AI.

Two tracks

One team, one standard, two directions.

APPLIED AI — BUILDING WITH A HOSPITAL PARTNER

Practical AI systems, built first for healthcare

Our first product is a Hospital AI System, built with a hospital partner: an AI receptionist that listens to patients and matches them with the right doctor, backed by a rule-based queue tracker. Healthcare is where we're focused — for now.

See the Hospital AI System →
RESEARCH — EARLY STAGE

A foundation language model for Bangla

We're fine-tuning open-weight LLMs on Bangla data so Bangladeshi businesses have a foundation to build on. Once the base model holds up, medical, marketing, and hiring versions follow.

Read about the initiative →

Track record

We've built AI systems before.

The same four of us run CodeAlignR, a US-based company delivering AI voice and chat automation since 2024. TaloSyn IT applies that same engineering discipline to healthcare, Bangla NLP, and whatever comes next.

More about the team

Since 2024

CodeAlignR

AI voice and chat automation for US clients, run by the team behind TaloSyn IT.

Product

Hospital AI System

An AI receptionist and scheduling system for hospitals — it listens to patients, matches them to the right doctor, and keeps queues honest in real time. Not everything here is AI, and that's deliberate: we use AI where judgment is genuinely needed and plain automation everywhere else — it's cheaper, faster, and easier for a hospital to trust.

  1. AI

    Listens and matches a doctor

    A patient calls or texts. The AI receptionist listens to the problem in Bangla or English and suggests the right doctor.

  2. Automated

    Books the appointment

    Once the patient confirms, the slot is booked automatically — no human scheduler involved.

  3. Automated

    Tracks the real queue

    A rule-based tracker counts how many patients are actually ahead of each appointment, in real time.

  4. AI

    Proposes a new time, doesn't impose one

    If the queue is running behind, the AI works out a revised time and sends it to the patient — and everyone booked after them — for confirmation before anything changes.

Status: Pilot stage.

Research

The Bangla LLM Initiative

We're adapting proven open-weight models — including Qwen and Gemma — into a language model that truly understands Bangla, dialects included, where most tools fall short.

Why

Most large language models treat Bangla as an afterthought. We're building one meant for it, so Bangladeshi businesses can build on it instead of working around it.

How

We fine-tune strong open-weight models on Bangla data rather than training from scratch — faster to a usable result, and light enough to run on hardware most companies can actually afford.

What's next

Once the base model holds up, we're tuning task-specific versions — starting with these areas:

Healthcare Social marketing Job search

Status: Early-stage research, not yet released.

About

Who's building this

TaloSyn IT is a four-person team based in Bangladesh. We also run CodeAlignR, delivering AI voice and chat automation to clients since 2024 — TaloSyn IT is where we build for Bangladesh directly.

Fahim Ibne Shams

Co-founder & Chief Product Officer

Fahim holds a B.S. in Computer Science from the University of North Texas and an M.S. from Harrisburg University. He built his industry experience at PNC and as a Full Stack Software Engineer at Plus One (an Optum Company), focusing on traditional backend systems and modern AI application engineering. He brings a strong technical enthusiasm for designing custom AI applications, voice agents, and RAG workflows. His vision centers on leveraging these automated architectures to solve practical, real-world infrastructure challenges at a localized level. He is a full-time Software Engineering Faculty member and Technical Mentor at Thaddeus Stevens College of Technology in Lancaster, Pennsylvania, where he teaches computer science and coaches teams for competitive programming competitions.

LinkedInlinkedin.com/in/fis

Kalyan Roy

Co-founder & Head of R&D

Kalyan Roy holds a Bachelor of Science in Computer Science and Engineering from North South University, Bangladesh. As an undergraduate, he was an active competitive programmer and competed in the ICPC Dhaka Regional. After graduating, he worked as a research assistant in the Department of Computer Science of North South University. His research has appeared at premier international venues: he co-authored a fast exact algorithm for enumerating maximal pseudo-cliques in large sparse graphs, published at KDD 2024, one of the world's top data mining conferences, and contributed to pioneering work on Bangla named entity recognition using deep learning, published in the Journal of Intelligent & Fuzzy Systems. He has received an offer of admission to the Master of Computer Science program at Carleton University in Ottawa, Canada. His research interests span algorithms, machine learning, and cryptography.

Websitekalyan.is-a.dev

Ramiza Maliha

Co-founder & Chief Technology Officer

Ramiza is a computer scientist who loves building things that are equal parts clever and useful, whether that's untangling the mysteries of single-cell biology or figuring out who the next Fernandinho is. She has worn both the instructor and researcher hats at North South University, and built her engineering chops as a Software Engineer at BJIT Limited, Bangladesh, spending two years turning coffee into working software for real clients. She's now pursuing her M.Sc. at Uppsala University, wrangling deep learning pipelines, building multimodal agentic AI chatbots, and debugging things that definitely worked five minutes ago by deadline.

LinkedInlinkedin.com/in/rami1423

Ifty Mohammad Rezwan

Co-founder & Chief AI Officer

Ifty is interested in AI-driven solutions and general solutions that make technology more interactive, engaging, and efficient.

He brings industry experience from multiple companies, working with top clients across computer vision, natural language processing, robotics, and efficient software solutions. He recently received a Ph.D. offer in AI in Education from the University of Central Florida, where he will focus on student misconceptions and AI pedagogy.

LinkedInlinkedin.com/in/ifty-mohammad-rezwan-98a487aa

Contact

Get in touch

Tell us which track this is about, and we'll route it to the right person.

Emailtalosyn.it@gmail.com

Phone / WhatsApp+8801749888998, +14696643921(Whatsapp)

Location House - 30, Road - 20, Sector - 13, Uttara, Dhaka