I didn't start with a grand plan. I just kept solving problems — for startups, for growing teams, for founders who needed someone to think alongside them, not just write code and disappear. Over the years, that turned into something real. I've built Android apps people actually use, data pipelines that process millions of events in real time, AI systems that make businesses feel 10x bigger than they are — and I've done it from scratch, with teams I handpicked, for clients across India, the Middle East, and Europe. Today I run Zentrix — a small, sharp engineering studio. We don't do bloated retainers or 40-slide proposals. We embed, we ship, we care about what happens after go-live. If you're building something that matters and need a technical partner who gets both the code and the business behind it — let's talk.

Subhadip Sasmal

I didn't start with a grand plan. I just kept solving problems — for startups, for growing teams, for founders who needed someone to think alongside them, not just write code and disappear. Over the years, that turned into something real. I've built Android apps people actually use, data pipelines that process millions of events in real time, AI systems that make businesses feel 10x bigger than they are — and I've done it from scratch, with teams I handpicked, for clients across India, the Middle East, and Europe. Today I run Zentrix — a small, sharp engineering studio. We don't do bloated retainers or 40-slide proposals. We embed, we ship, we care about what happens after go-live. If you're building something that matters and need a technical partner who gets both the code and the business behind it — let's talk.

Available to hire

I didn’t start with a grand plan. I just kept solving problems — for startups, for growing teams, for founders who needed someone to think alongside them, not just write code and disappear.
Over the years, that turned into something real. I’ve built Android apps people actually use, data pipelines that process millions of events in real time, AI systems that make businesses feel 10x bigger than they are — and I’ve done it from scratch, with teams I handpicked, for clients across India, the Middle East, and Europe.
Today I run Zentrix — a small, sharp engineering studio. We don’t do bloated retainers or 40-slide proposals. We embed, we ship, we care about what happens after go-live.
If you’re building something that matters and need a technical partner who gets both the code and the business behind it — let’s talk.

See more

Experience Level

Expert
Expert
Expert
Expert
Expert
Expert
Expert
Expert
Intermediate
Intermediate
Intermediate
Intermediate
See more

Language

Bengali
Fluent
English
Fluent
Abkhaz
Intermediate
Chinese
Advanced
Hindi
Fluent

Work Experience

Multilingual Language Expert at Mercor
June 1, 2025 - August 1, 2025
Evaluated AI-generated content in Bengali for fluency, clarity, and grammatical correctness. Identified linguistic errors and provided feedback on cultural context and natural language usage patterns. Reviewed and annotated responses following structured guidelines to ensure consistency and quality standards.
AI Training Expert at Outlier.ai
February 1, 2025 - May 1, 2025
Fine-tuned large language models using RLHF, improving response quality and contextual accuracy by 18%. Curated and annotated conversational datasets for intent classification, sentiment analysis, and safety filtering. Collaborated with NLP engineers to design automated evaluation pipelines and prompt libraries for rigorous benchmarking and rapid iteration.
Data AI Engineer at v4cai
June 1, 2025 - Present
Engineered and optimized scalable ETL pipelines using Python and SQL, reducing data processing latency by 40% for real-time model inference. Developed Retrieval-Augmented Generation (RAG) systems to enhance LLM accuracy, integrating vector databases for semantic search and context retrieval. Deployed and monitored machine learning models in production environments, ensuring 99.9% uptime and optimizing resource utilization through quantization techniques.
Data Scientist Intern at Fractal Analytics
August 1, 2024 - October 1, 2024
Trainee at Salesforce Virtual Internship
May 1, 2024 - July 1, 2024
LLM agentic trainer at Turing
February 28, 2026 - October 28, 2025

Education

B.Tech in Computer Science at Kalinga Institute of Industrial Technology (KIIT University)
May 20, 2021 - September 26, 2025

Qualifications

Cisco CCNA Certification
January 11, 2030 - February 20, 2026
Salesforce Virtual Internship Completion Certificate
January 11, 2030 - February 28, 2026
AWS Certified Solutions Architect
March 3, 2025 - March 30, 2026
Learned AWS multi-tier architecture, VPC setup, EC2, RDS, S3, CloudFront, IAM, and built scalable, fault-tolerant systems with cost optimization.
Azure Administrator (AZ-104)
March 30, 2026 - March 30, 2026

Industry Experience

Education, Professional Services, Software & Internet, Computers & Electronics
    paper verse.ai — Multi-Tenant RAG Platform. check it http://3.23.132.183:3001/

    Most companies are sitting on a goldmine of knowledge — internal docs, handbooks, reports, wikis — that nobody can actually find when they need it. Employees spend hours digging through folders, pinging the one person who “knows everything”, or just guessing.
    verse.ai fixes that.
    It’s a multi-tenant AI knowledge platform built for organizations that want their people to actually use the knowledge they’ve spent years building. Each company gets its own private workspace — their docs go in, their team asks questions in plain English, and the AI answers back with exact, cited answers pulled only from their data. Not some generic model hallucinating on the internet. Their actual documents.
    The platform is built around three simple roles: a super admin who provisions organizations, a tenant admin who uploads the knowledge base, and team members who just… ask. No training required. No prompt engineering. Just a chat box and an answer.
    Under the hood, every org’s data lives in a fully isolated vector namespace — Acme Corp can never see Globex’s files, and vice versa. It’s not just a permission flag, it’s architectural isolation. The AI literally has no path to cross-tenant data.
    Built it for a Brazilian client who needed this running across multiple organizations simultaneously — and it delivers sub-50ms responses with 99% uptime.
    If your team is still answering the same question 40 times a week because nobody can find the doc that answers it — this is what you build.