Experience
Software Developer (Officer Grade-I), IT Division
- Organization
- Bangladesh Finance PLC
- Location
- Dhaka, Bangladesh
- Period
- Sep 2025 – Present
- Type
- Full-time
Designing and building the in-house systems behind Bangladesh Finance PLC's paperless, green, and cyber-secured digital transformation, from customer onboarding and document workflows to security operations.
Responsibilities
- Architected and built AEGIS, an open-source security operations platform on Wazuh that unifies endpoint, firewall, and Microsoft 365 telemetry with explainable incident correlation and risk scoring
- Contributed to FinX360, the company's digital ecosystem fronting 20 applications, including DigiTion, DGdocX, PMS, and the Lead Management System
- Built the eKYC platform (FastAPI, AES-256-GCM encryption, JWT/RBAC, Election Commission NID API, OCR, maker-checker workflow) and designed a shared eKYC hub for three group entities with BFIU compliance in mind
- Built the Smart Audit Management System, GRC, DGServices centralized authentication, and IT asset management
- Leading backend development for FINIX (Admin and Customer Portals), a Phishing Automation System for the October Cyber Security Awareness program, and a Meeting Minutes Generator
- Taking on Oracle DBA ownership of the Core Banking System (CBS)
- Produced technical documentation and architecture diagrams for the audit committee
Achievements
- Enabled 60% faster audit cycles through rule-based audit automation
- Delivered a production-ready ICT risk inspection tool used across branches
- Took AEGIS from a single practice VM to a live office pilot ingesting four telemetry sources
Stack: FastAPI, Python, Go, PostgreSQL, Oracle, Flutter, React, Next.js, Wazuh, YARA, Celery, RabbitMQ, LangChain, Groq, RAG Pipelines, JWT Auth
Software Engineer (Intern)
- Organization
- Bangladesh Finance PLC
- Location
- Dhaka, Bangladesh
- Period
- Apr 2025 – Aug 2025
- Type
- Internship
Contributed to enterprise audit automation, internal inspection tools, and secure customer onboarding modules using AI-assisted workflows.
Responsibilities
- Helped build the Smart Audit Management System with automated rule-based findings from policy documents
- Enhanced the Lead Management System with OTP login, role control, and high-security activity logging
- Developed eKYC modules including cloth detection, OCR-based NID extraction, and liveness verification
- Supported backend architecture improvements for enterprise-grade reliability and security
Achievements
- Played a key role in rolling out audit and inspection automation modules to production
- Built foundational modules for eKYC and security tools used by ICT and Operations teams
Stack: FastAPI, Flutter, PostgreSQL, LangChain, Groq, ML Kit, pytesseract
Research Intern
- Organization
- Center for Computational & Data Sciences (CCDS), IUB
- Location
- Dhaka, Bangladesh
- Period
- Jan 2025 – Present
- Type
- Research
Conducting research on NLP, LLMs, and educational AI with a focus on Bangla and AI-powered tutoring systems.
Responsibilities
- Developed the SOMAJGYAAN dataset for evaluating LLMs on Bangla culture and social knowledge
- Participated in the BEA 2025 and BLP-2025 shared tasks on educational feedback and code generation
- Ran user studies on PACE, an LLM-based Python tutor, analyzing engagement and learning outcomes
- Co-authored papers at CHI EA 2025, BEA 2025, and IJCNLP-AACL 2025
Achievements
- First author on the CHI 2025 Extended Abstract and co-author on four further publications
- Helped introduce one of the first Bangla cultural reasoning benchmarks
- Top-3 team in the BEA 2025 shared task
Stack: Python, PyTorch, Hugging Face, LangChain, Pandas
Undergraduate Researcher
- Organization
- Center for Computational & Data Sciences (CCDS), IUB
- Location
- Dhaka, Bangladesh
- Period
- Sep 2023 – Dec 2024
- Type
- Research
Undergraduate thesis on LLM-powered Python tutoring and human-AI interaction in learning environments.
Responsibilities
- Designed and implemented the PACE tutoring system architecture
- Conducted a user study to evaluate engagement and learning gains
- Analyzed quantitative and qualitative metrics to understand tutor effectiveness
- Prepared the manuscript that led to the CHI 2025 Extended Abstracts publication
Achievements
- Observed about 35% improvement in learning outcomes in controlled study settings
- Selected for SIGCHI Winter School 2024 to present a position paper
Stack: Python, LangChain, FastAPI, React, PostgreSQL
Junior ML Engineer
- Organization
- Omdena Liberia Chapter
- Location
- Remote
- Period
- Jan 2024 – Jul 2024
- Type
- Volunteer
Built machine learning models for malaria risk prediction in a global, impact-driven collaboration.
Responsibilities
- Engineered predictive models for malaria outbreak risk using tabular health and climate data
- Preprocessed and analyzed epidemiological datasets to improve data quality
- Collaborated with a global team of 20+ ML engineers in a fully remote, agile setup
Stack: Python, Scikit-learn, Pandas, XGBoost
Publications
Google Scholar: 32 citations, h-index 2, i10-index 2
SOMAJGYAAN: A Dataset for Evaluating LLMs on Bangla Culture, Social Knowledge, and Low-Resource Language Adaptation
Fariha Anjum Shifa, Muhtasim Ibteda Shochcho, Abdullah Ibne Hanif Arean, Mohammad Ashfaq Ur Rahman, Akm Moshiur Rahman Mazumder, Ahaj Mahhin Faiak, Md Fahim, M Ashraful Amin, Amin Ahsan Ali, AKM Mahbubur Rahman
- Venue
- Proceedings of the 14th International Joint Conference on Natural Language Processing and the Asia-Pacific Chapter of the ACL (IJCNLP-AACL 2025)
- Year
- 2025
- Type
- Conference
- Abstract
- A benchmark for evaluating large language models on Bangla cultural knowledge, social commonsense, and low-resource linguistic nuance.
- My contribution
- Dataset design, evaluation framework, analysis.
- Impact
- Introduces a large cultural reasoning benchmark for Bangla.
- Tags
- Dataset, Low-Resource NLP, Cultural Reasoning, LLM Evaluation
Improving User Engagement and Learning Outcomes in LLM-Based Python Tutor: A Study of PACE
Muhtasim Ibteda Shochcho, Mohammad Ashfaq Ur Rahman, Shadman Rohan, Ashraful Islam, Hasnain Heickal, AKM Mahbubur Rahman, M Ashraful Amin, Amin Ahsan Ali
- Venue
- Extended Abstracts of the CHI Conference on Human Factors in Computing Systems (CHI EA 2025)
- Year
- 2025
- Type
- Extended Abstract
- Citations
- 10
- Abstract
- A controlled user study on the PACE tutoring system showing improvements in engagement and conceptual understanding through LLM-based guidance.
- My contribution
- Lead researcher. System design, experiment setup, data collection, statistical analysis.
- Impact
- About 35% improvement in learning outcomes for introductory Python learners.
- Tags
- HCI, EdTech, User Study, LLM
BD at BEA 2025 Shared Task: MPNet Ensembles for Pedagogical Mistake Identification and Localization in AI Tutor Responses
Shadman Rohan, Ishita Sur Apan, Muhtasim Ibteda Shochcho, Md Fahim, Mohammad Ashfaq Ur Rahman, AKM Mahbubur Rahman, Amin Ahsan Ali
- Venue
- 20th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2025)
- Year
- 2025
- Type
- Workshop
- Abstract
- An MPNet-based ensemble for detecting and localizing pedagogical mistakes in AI-generated tutoring responses.
- My contribution
- Model architecture design, ensemble strategy, evaluation.
- Impact
- Top-3 performance among all teams in the shared task.
- Tags
- Educational NLP, Error Detection, Ensemble Models, LLM Evaluation
AlphaBorno at BLP-2025 Task 2: Code Generation with Structured Prompts and Execution Feedback
Mohammad Ashfaq Ur Rahman, Muhtasim Ibteda Shochcho, Md Fahim
- Venue
- Second Workshop on Bangla Language Processing (BLP-2025), IJCNLP-AACL 2025
- Year
- 2025
- Type
- Workshop
- Abstract
- A structured prompting strategy combined with execution feedback loops to improve code generation from Bangla instructions.
- My contribution
- Prompt engineering, evaluation pipeline, error analysis.
- Impact
- A structured code generation pipeline from Bangla instructions.
- Tags
- Code Generation, LLMs, Bangla NLP, Prompting
Fruit Quality Detection and Monitoring System
Md Maruf Kamran Sohag, Mohammad Ashfaq Ur Rahman, Muhtasim Ibteda Shochcho, Md Ridwan Mahmud, Daiyan Mohammad Shams, Mysha Samiha, Mohammad Rejwan Uddin, Mahady Hasan
- Venue
- 2023 IEEE 13th International Conference on Control System, Computing and Engineering (ICCSCE)
- Year
- 2023
- Type
- Conference
- Citations
- 13
- Abstract
- An IoT-assisted computer vision framework for real-time fruit quality detection and automated classification.
- My contribution
- Real-time IoT-based fruit quality detection and monitoring system.
- Impact
- End-to-end automated pipeline for fruit quality monitoring.
- Tags
- Computer Vision, IoT, Agritech
Thesis: PACE: Python AI Companion for Enhanced Engagement
Muhtasim Ibteda Shochcho, Mohammad Ashfaq Ur Rahman
- Venue
- Independent University, Bangladesh
- Year
- 2024
- Type
- Undergraduate Thesis
Official projects at Bangladesh Finance PLC
AEGIS: Security Operations Platform
- Status
- Pilot
- Role
- Architect & Lead Developer
- Period
- July 2026 – Present
- Category
- Cybersecurity
An open-source security operations platform built on Wazuh for Bangladesh Finance PLC. AEGIS brings endpoint telemetry, YARA malware scanning, two perimeter firewalls, and Microsoft 365 audit logs into one pipeline, then turns raw alerts into scored, explainable incidents an analyst can act on.
Highlights
- Ingests four live telemetry sources: Wazuh/Windows endpoint events, YARA file scans, FortiGate and Palo Alto firewall logs, and Microsoft 365 sign-in audit events
- Built the Fusion Triage Engine, which normalizes every source into a unified security record, persists it to an append-only PostgreSQL ledger, and runs every five minutes
- Correlation engine groups related events into incident candidates by user session, filtering out machine and service accounts
- Versioned, replayable risk scoring (0-100) weighs activity diversity, source diversity, severity, and burst compression, with damping for high-volume noise; high-risk incidents are written back to the Wazuh dashboard
- Explainable by construction, with no LLM or GPU dependency
- High-level design mapped to the Bangladesh Bank ICT Security Guideline v4.0
Impact: Processed hundreds of thousands of events in the office pilot and surfaced real, previously unseen security findings for the network team.
Stack: Wazuh, YARA, Python, PostgreSQL, FortiGate syslog, Palo Alto PAN-OS, Microsoft 365 Management API, systemd, Linux
FinX360 Digital Ecosystem
- Status
- Live
- Role
- Developer, IT Division team
- Period
- September 2025 – Present
- Category
- Enterprise & FinTech
Bangladesh Finance PLC's in-house digital ecosystem, a single portal fronting 20 applications (13 internal systems and 7 external links). It began during COVID as a business continuity tool and grew into the backbone of the company's paperless operations.
Modules
- DigiTion
- approvals, digital memos, DVR, loan and EMI monitoring, NID verification, case management
- DGdocX
- QR-based physical document tracking with check-in and check-out
- PMS
- KPI/KRA and appraisal lifecycle
- LMS
- lead management for Relationship Managers
Impact: Core of the company's paper-to-paperless transformation.
Stack: FastAPI, PostgreSQL, React, Flutter
eKYC Platform & Shared eKYC Hub
- Status
- Live
- Role
- Full-Stack Developer
- Period
- September 2025 – Present
- Category
- FinTech & Computer Vision
Mobile and backend KYC solution with face verification, liveness detection, OCR-based NID extraction, and document validation, extended into a shared eKYC hub architecture for three group entities (BFL, BFCL, BFSL).
Highlights
- AES-256-GCM encryption for customer data, JWT with role-based access control
- Election Commission NID API integration and maker-checker approval workflow
- Real-time liveness detection to prevent spoofing, plus a cloth detection model for dress code compliance
- Designed the shared hub with Bangladesh Financial Intelligence Unit (BFIU) compliance in mind
- Integrating biometric fingerprint capture into the onboarding flow
Impact: Onboarded 10,000+ customers with a 3-minute average KYC time.
Stack: Flutter, FastAPI, ML Kit, EasyOCR, Groq, Celery, RabbitMQ, PostgreSQL, Firebase
Smart Audit Management System
- Status
- Live
- Role
- Developer
- Period
- April 2025 – present
- Category
- Enterprise & FinTech
Compliance and audit management platform with rule-based findings from policy documents, risk assessment, role-based access, analytics, and automated audit trails.
Highlights
- Automated rule-based findings across multiple policy documents
- Role-based access control for auditors and reviewers
- Analytics dashboard and full audit trail
Impact: 60% faster audit cycles; 500+ audit documents processed monthly.
Stack: FastAPI, PostgreSQL, React
Smart ICT Risk Inspection Tool
- Status
- Live
- Role
- Developer
- Period
- October 2025
- Category
- Cybersecurity & Compliance
Structured ICT risk inspection with scoring, reporting, and automated inspection summaries.
Impact: Used across branches for ICT inspections.
Stack: FastAPI, PostgreSQL, React
GRC Platform
- Status
- Built, awaiting launch
- Role
- Developer
- Period
- September 2025 – present
- Category
- Cybersecurity & Compliance
Governance, risk, and compliance system supporting the company's regulatory and internal control processes.
Stack: FastAPI, PostgreSQL, React
DGServices: Centralized Authentication
- Status
- In development
- Role
- Developer
- Period
- 2026
- Category
- Security & Identity
Single sign-on and centralized authentication service shared across the company's internal applications.
Stack: FastAPI, PostgreSQL, JWT
Phishing Automation System
- Status
- In development
- Role
- Lead Developer
- Period
- 2026
- Category
- Cybersecurity
In-house phishing simulation platform built from scratch for the October Cyber Security Awareness program. Designed as an employee-friendly awareness assessment rather than a gotcha.
Highlights
- Campaign setup with link and attachment phishing scenarios
- Post-click reveal experience that teaches in the moment
- One-click phishing reporting and gamification
- IT-monitored dashboard and governance controls
FINIX Super App (Admin & Customer Portals)
- Status
- In development
- Role
- Backend Developer
- Period
- January 2026 – Present
- Category
- FinTech
Backend for the company's customer-facing super app, comprising an Admin Portal and a Customer Portal.
Meeting Minutes Generator
- Status
- In development
- Role
- Backend Developer
- Period
- March 2026 – Present
- Category
- AI & NLP Systems
Backend for automated generation of structured meeting minutes.
Lead Management System
- Status
- Live
- Role
- Backend Developer
- Period
- June 2025 – 2026
- Category
- Enterprise CRM
CRM backend (the LMS module of FinX360) with OTP authentication, audit logging, and lead pipeline management.
Highlights
- OTP authentication and role-based access
- Comprehensive audit logging
- Lead scoring, pipeline automation, email and SMS notifications
Impact: Managing 10,000+ leads with a 40% conversion improvement.
Stack: FastAPI, PostgreSQL, Redis, OAuth, React
Asset Management Systems
- Status
- Live
- Role
- Backend Developer
- Period
- May 2025 - 2026
- Category
- Enterprise Systems
IT asset tracking with risk and criticality scoring and lifecycle management, plus a Fixed Asset Management Module.
Impact: 200+ assets tracked across departments.
Stack: FastAPI, PostgreSQL, React, TypeScript
XBroker: AI-Powered Broker Portfolio Analyzer
- Status
- Live
- Role
- Lead Developer
- Period
- June 2025
- Category
- AI & NLP Systems
PDF ingestion and AI summarization system that generates structured broker profiles and insights.
Highlights
- PDF parsing pipeline for unstructured broker documents
- RAG-based portfolio summarization into structured JSON profiles
Impact: 500+ broker portfolios processed; manual profile work reduced by 80%.
Stack: Python, FastAPI, LangChain, PostgreSQL, RAG, OpenAI
Personal, client and research projects
PACE: Python AI Companion for Enhanced Learning
- Status
- Published (CHI EA 2025)
- Role
- Researcher
- Period
- 2023 – 2024
- Category
- AI in Education
LLM-powered intelligent tutoring system for Python programming with adaptive feedback and misconception detection.
Highlights
- Adaptive tutoring with personalized learning paths
- Real-time code analysis and feedback generation
- Controlled user study published at CHI 2025 Extended Abstracts
Impact: About 35% improvement in learning outcomes compared to traditional methods.
Stack: Python, LangChain, FastAPI, React, PostgreSQL, OpenAI API
Telemedicine & Healthcare Consultation Platform
- Status
- In development
- Role
- Technical Lead (client project)
- Period
- 2026 – Present
- Category
- HealthTech
Real-time healthcare platform for the Bangladesh market with video consultations, e-prescriptions, an online pharmacy, and subscription health packages. Leading a four-engineer team across backend, Flutter mobile, and a Next.js admin panel.
Highlights
- Separate patient and doctor apps covering booking, waiting room, live video, in-call chat, and prescriptions
- Order service designed as a saga state machine with a transactional outbox and Kafka-driven notifications
- Hosted-checkout payment integration with idempotent, independently verified webhooks
- Prescription-to-pharmacy flow that converts dosage and duration into orderable quantities
Stack: Go, Flutter (BLoC), Next.js, PostgreSQL, Kafka, Agora, Firebase Cloud Messaging, PayStation
NETSCOPE: Animated Network Traffic Visualizer
- Status
- In design
- Role
- Designer & Developer
- Period
- August 2026 – Present
- Category
- Cybersecurity & Visualization
"Wireshark, but animated." A Matrix-style visualizer that renders IP-to-IP traffic on a LAN in real time, driven by firewall logs instead of packet capture.
Highlights
- Builds on the normalized security records produced by AEGIS
- Strict performance budgets and a six-phase build plan
Stack: Go, React, Tailwind CSS, Canvas, PostgreSQL
Malaria Risk Prediction (Omdena)
- Status
- Completed
- Role
- Junior ML Engineer
- Period
- 2024
- Category
- ML for Social Good
Predictive models for malaria outbreak risk using health and climate data, built with a global volunteer team.
Stack: Python, Scikit-learn, XGBoost, Pandas