AI Solutions & Product Builder
I build systems that turn messy inputs into clear, usable outputs.
My freelance work spans AI agents, ML sensor analytics, automation for smaller businesses and custom CRMs. I connect requirements, solution design and implementation, alongside building BricDesk for real estate operations.
I work best in environments where I can help shape the structure of a problem, then build a practical system or prototype around it.
What I Build
Freelance — AI Solutions & Application Development
- Built task-focused AI agents and workflow automation for MSMEs, integrating paid OpenAI and Google Gemini services with business tools.
- Built organization-specific CRMs, client websites and applications around stakeholder requirements and operational workflows.
- Built a residential solar configuration app connecting property dimensions and electricity needs with panel options, visual layouts and estimated costs.
- Contributed to sensor calibration and anomaly detection for confidential plastic bottle-cap and Pune automobile-parts manufacturing clients.
Client identities remain confidential. Freelance scope and contributions
BricDesk — Independent Product Development
Building a real estate operations product connecting inventory, lead and channel partner ownership, holds, approvals, and founder visibility.
- Work across workflow discovery, Flutter application development, Supabase-backed data flows, and responsive web experiences
- Focus on explicit role and project context, accountable actions, and product iteration
- Complement product work with client websites, applications, AI automation, and tool integrations
Independent AI & Application Implementations
Built an AI assistant that compares machine readings, shows its sources and drafts a note for human review. A separate model learns leakage patterns from public research data; its test results and missed cases are documented.
Demonstration tooling: local Qwen, LangChain/Ollama and Python/NumPy. The classifier uses 30 sensor features; 552 test recordings from the same rig include 15 missed cases.
Trained a support-request classifier and evaluated it on 3,080 held-out messages. Compared keyword, embedding and hybrid search across 5,183 abstracts and 300 queries, retaining results and failure cases.
Application work includes interactive solar configuration, supplier comparison, shift planning and source-linked website-support handovers.
These are current independent implementations, separate from client delivery. NLP studies · Industrial AI workflow · Working applications
How I Think
Most messy problems become easier once the structure is clearer.
• Breaking problems into systems• Simplifying workflows• Designing for real conditions
Experience
Tata Technologies - GenAI Intern
Worked on applied AI workflow concepts in an enterprise setting.
- Explored agentic and multilingual workflow concepts
- Learned how operational fit shapes system design
Skills
Solution discovery, workflow design, OpenAI/Gemini API integration, Microsoft-provided enterprise AI services, LangChain/Ollama demonstration tooling, supervised NLP, TF-IDF, scikit-learn, sentence embeddings, ONNX Runtime, ML evaluation, source-grounded handovers, product implementation.
Cloud exposure: Introductory familiarity with AWS and Microsoft Azure.
Education
B.Tech - Computer Science
MIT-WPU, Pune