Job Description – Data Scientist / Applied ML Engineer 

Location: August Kranti Marg, Siri Fort Institutional Area, near Siri Fort Auditorium, New Delhi - 110049 

Experience: 4 – 6 years 

C2H - 3 months with the possibility of an extension based on business requirements.

Budget - 75k to 80K per month

Role Overview 

We are looking for a highly skilled Data Scientist / Applied ML Engineer to build and optimize DigiCatalog’s probabilistic AI stack for catalogue intelligence, multilingual understanding, and product attribute extraction. 

The role focuses on selecting, orchestrating, and evaluating open-source AI models across OCR, computer vision, multimodal AI, image enhancement, and Indic language processing. The ideal candidate should have strong expertise in applied CV/NLP systems, model evaluation, pipeline optimization, and practical ML engineering with a strong focus on cost, latency, and quality trade-offs. 

This role emphasizes building efficient, purpose-built AI pipelines rather than relying on large general-purpose models. 

Key Responsibilities 

Design and orchestrate AI/ML pipelines using open-source models across:  

OCR  

Computer Vision  

Multimodal AI  

Indic language processing  

Image enhancement and correction  

Select and optimize models based on:  

Cost  

Latency  

Accuracy  

Deployment constraints  

Edge-to-cloud trade-offs  

Build probabilistic AI workflows using small specialized models instead of defaulting to large LLMs.  

Work with technologies and frameworks such as:  

DocTR  

PaddleOCR  

TrOCR-class models  

Grounding DINO  

OWL-ViT  

CLIP  

2–3B parameter VLMs  

Real-ESRGAN  

SDXL-Turbo class models  

IndicTrans2  

NLLB  

IndicWhisper  

Build and maintain:  

Golden datasets  

Automated evaluation pipelines  

Regression testing frameworks  

Model benchmarking systems  

Define and monitor evaluation metrics including:  

Attribute F1 score  

OCR CER/WER  

Hallucination rate  

Per-language performance slices  

Perform:  

Error analysis  

Threshold tuning  

Calibration  

Ensembling  

Post-processing optimization  

Run regression validation on every model replacement or pipeline change.  

Clean, curate, and prepare datasets for training and evaluation workflows.  

Fine-tune models only when justified through measurable cost and quality improvements.  

Continuously evaluate emerging research and rapidly productionize relevant open-source innovations.  

4-6 years Immediate C2H joiners 

Required Skills & Qualifications 

4–6 years of experience in:  

Applied Machine Learning  

Computer Vision  

NLP  

Multimodal AI systems  

Strong hands-on expertise in:  

PyTorch  

Python  

ML pipeline orchestration  

Model evaluation systems  

Experience with:  

OCR systems  

Vision-language models  

Image processing pipelines