Abhijith C M

Software Engineer (AI/ML) — developer & owner of Chanakya

Bengaluru, India chanakya-finance.com

Same document as the page below — kept in one place so they never diverge.

Résumé

Summary

Software engineer with 3.5+ years in software development, spent in FinTech (BFSI) building AI/ML systems that read financial and KYC documents: classification, data extraction, document understanding, and validating what a customer claims against source data. Also responsible for deploying and running the Python backend services behind that work.

Experience

Software Engineer (AI/ML)

Decimal Technologies Private Limited · Bengaluru

Feb 2025 — Jul 2026

Decisioning and validation of customers from the information and the KYC and financial documents they provide, analysing the documents and checking them against source data through internal and third-party APIs.
Deploying and managing all Python backend services and their logs.

Software Engineer (AI/ML)

Namaste Credit · Bengaluru

Sep 2022 — Jan 2025

Over two years in FinTech, researching, developing and integrating classification, data extraction and document understanding for financial and KYC documents.
Work centred on raising the accuracy and the efficiency of those processes.

Engineering Intern

Operative India Private Limited · Bengaluru

Mar 2022 — Aug 2022

QA engineering: automated testing with scripts and generated reports through Jenkins.

Projects

VahanaAI

VLMs · vLLM · Docker

Feb 2025 — Present

Model pipelines for structured data extraction from banking and financial documents, through to deployment.
Research on signature matching and document forgery detection.

FTR / Gatekeeping

Data analysis · Flask · MySQL

Feb 2025 — Present

Source data collection: integrated internal document extraction and analysis APIs, plus Vahana Hub wrapper APIs for sources such as GST, Udyam, NSDL PAN and UIDAI.
Validation: compared source data against what the customer provided — name, date of birth, address, and business details for a business.
Decisioning: a report combining the KYC, financial, source and customer-provided comparisons, driven by those validations.

KYC document classification, data extraction and masking

OpenCV · YOLO · OCR · PyTorch · Transformers · Flask · Celery · RabbitMQ · Redis

Sep 2022 — Jan 2025

Object detection: YOLO models classifying PAN, Aadhaar, driving licence, passport and voter ID, and locating the key information areas to extract.
OCR: tested and integrated open-source OCR models to improve extraction, plus a model for detecting document orientation.
OCR-free understanding: trained and integrated the DONUT multimodal model.
Asynchronous processing: a Celery worker for the masking pipeline.
Accuracy and speed: continuous retraining on new data, extra cleaning steps, and code optimisation to cut processing time.

Banking and financial data extraction and analysis

YOLO · Mask R-CNN · Scikit-Learn · PyTorch · Pandas · Ghostscript · BeautifulSoup · Llama 2 · AWQ · vLLM

Apr 2023 — Jan 2025

Built a bank-logo dataset by web scraping, for classifying bank statements.
Trained a model extracting basic details from bank statements and contributed to tabular data extraction; extended the same to GST, ITR and Udyam documents.
Trained ML models and a BERT model to classify banking transactions for analysis.
Cut inference time substantially with quantization and vLLM on a fine-tuned LLM used for named entity recognition.

Skills

Languages: Python
Machine learning: Deep learning NLP CNNs LLMs VLMs Computer vision Document understanding OCR Multimodal models
Frameworks: PyTorch TensorFlow OpenCV Scikit-Learn Transformers Flask FastAPI Celery Streamlit
Data & infra: MySQL MongoDB RabbitMQ Redis Docker vLLM AWQ quantization Ghostscript Linux

Education

BE, Computer Science and Engineering

BMS Institute of Technology and Management · Bengaluru, Karnataka · CGPA 7.77

Aug 2018 — Jul 2022