Freelance · India based · SGT overlap
Chatbots, retrieval-augmented generation, and LLM-based automation that answer from your own documents, tested against known-answer questions before launch, not a generic demo.
Plain answer
A working production system, not a notebook, prototype, or deck that your team still needs to finish.
A freelance generative AI developer designs a RAG pipeline or fine-tuning approach so the system answers from your own data, tests it against known-answer questions before launch, and deploys it with monitoring for hallucination rate and output drift. For a Singapore business, that also means answering to IMDA's generative AI governance framework alongside PDPA.
Inside the engagement
The person who scopes the project is the same person building each layer, so fewer handoffs and clearer decisions.
Every deliverable connects to the next, so validation assumptions do not get lost between specialists.
Your documents audited for retrieval readiness before the architecture is built.
Prompt-and-retrieval or fine-tuning, chosen based on your data volume and control needs.
Outputs checked against known-answer cases before launch, then monitored as the underlying model provider updates its API.
What I build
Generative AI development is one of six areas of AI and ML work. The person who scopes your project is the person who builds it, across all six.
Churn prediction, demand forecasting, and risk scoring, validated on your actual data using Python, TensorFlow, PyTorch, and scikit-learn.
NLPDocument processing, feedback classification, and text intelligence built with BERT, spaCy, and transformer-based models.
VisionMedical imaging analysis, quality control, and retail automation, trained on your specific image domain, cloud or edge inference.
Custom chatbots, RAG pipelines, and LLM fine-tuning powered by GPT-4, Claude, Gemini, and open-source models.
Autonomous agents that plan, reason, and take action using LangChain, AutoGen, and CrewAI for multi-step workflows.
TrainingFine-tuning or from-scratch training on your proprietary data, with versioning and production deployment.
Compare the paths
See the communication path and commitment behind three common hiring options. Singapore's National AI Strategy 2.0 aims to triple the national AI talent pool from about 4,500 to 15,000 professionals by 2027, which is part of why local AI hiring runs slow and expensive right now.
| Option | Typical cost | Time to start | Commitment |
|---|---|---|---|
| Full-time in-house | S$100,000-170,000 base (Morgan McKinley, 2026), plus CPF and benefits | Weeks to months | Ongoing headcount |
| AI development agency | Typically 2-3x a freelance developer's base cost | Days to weeks | Team capacity, higher overhead |
| Freelance AI developer | USD 50-150/hr or USD 2,000-50,000 per project | Starts within 48 hours | Scales per project |
Compliance architecture
If the system retrieves or generates text involving personal data, PDPA's consent and purpose-limitation rules still apply. What is specific to generative AI is IMDA's 2024 framework: documented hallucination-testing, a bias check on generated content, and a clear position on who owns AI-generated output. Singapore's PDPA carries penalties of up to S$1 million or 10% of annual turnover, whichever is higher, under the 2020 amendment.
This is background information, not legal advice. Confirm your organisation's specific obligations with a qualified data protection officer or lawyer.
Transparent delivery
Interim reviews keep business assumptions, technical choices and compliance decisions visible.
A 30-minute call during the Singapore-India overlap covering what the system needs to answer and where its source data lives before anything is quoted.
Your documents or data get audited for retrieval readiness, and the architecture gets mapped against your accuracy and governance requirements.
The system is built and tested against known-answer questions and edge cases, with interim check-ins rather than a single reveal at the end.
The system goes live on your infrastructure, with monitoring for hallucination rate and output drift, not a one-time handoff.
Working rhythm
A Singapore morning call lands in late morning in India, while most afternoon reviews still fit the India workday.
Flexible engagement
Tooling
No mandatory framework. The choices follow your data, infrastructure and operational constraints.
Providers and frameworks chosen for your accuracy and control needs.
Documents made retrievable with the right chunking and search strategy.
Hallucination rate and drift treated as an ongoing operational metric.
Proof over promises
Twelve case studies across machine learning, NLP, computer vision and generative AI.
Custom entity extraction and scoring trained on the client's own criteria.
“Understood our requirements well and built a solution that saved a lot of manual work.” Utsav, CEO
See the problem, approach and measured result across production AI projects.
Client testimonials
Real feedback from clients who hired an AI and ML developer for automation and custom ML solutions.
"Working with Shreyans was a smooth experience. He understood our requirements well and built a Credit Report Automation solution that saved us a lot of manual work and made our process much faster."
"Shreyans quickly understood our process and delivered an automation solution that made audit form filling much easier for our team. It reduced manual work and improved overall efficiency."
"Shreyans delivered an AI chatbot that handled our most common customer queries automatically. Response times dropped significantly and the solution was well documented and straightforward to maintain."
"The NLP pipeline Shreyans built cut our data processing time by 70%. Outstanding technical depth and great communication throughout."
"The computer vision solution Shreyans built reduced defect detection time by 80% on our manufacturing line. Exceptional work."
"The generative AI system Shreyans built saved us 30 hours per week in content creation. Highly recommend for any AI project."
Why work with me
Specialising in generative AI, machine learning, LLM integration, and predictive analytics, systems are built secure, scalable, and high-performance, for real-world business impact, not experimental demos.
A clear, honest evaluation of your business needs and data, focused on AI solutions that are realistic, scalable, and built to create real impact.
Every solution is built around your business needs. Whether generative AI, LLM integration, or machine learning models, the goal is to solve real problems.
Robust, secure, scalable AI systems designed and deployed to adapt, perform, and grow alongside your business.
From planning to deployment, the full process is managed to make sure everything works properly and keeps improving over time.
Industries we serve
Domain expertise across diverse sectors, built to solve real business problems, not adapted from a generic template.
About
I build intelligent AI and ML solutions that help businesses solve real problems and make smarter decisions: machine learning models, deep learning, NLP and computer vision, data analysis and insights, and AI-powered automation.
Blog
Notes on shipping production AI, written from the build, not the brochure.
The move from general AI to vertical LLMs and autonomous agents driving faster, industry-focused decisions.
02 May 2026 Data scienceHow ML consulting helps businesses turn raw data into actionable insights and sharper strategy.
01 May 2026 Machine learningHow AI extracts data and reduces manual work in insurance document processing, with key use cases.
30 Apr 2026FAQ
Clear answers on cost, compliance, location and timing.
Start with the data
Bring the documents or data the system needs to answer from, and any governance requirements already on your plate. The first call is 30 minutes, timed inside the Singapore-India overlap.
