I engineer production-grade software systems at the intersection of full-stack
development and applied AI — from multi-agent LLM orchestration and retrieval-
augmented generation pipelines to real-time full-stack platforms serving live
video, data, and CRM workflows.
My work is grounded in three principles: systems should be verifiable, systems
should be observable, and systems should be built for scale from day one.
Core focus areas:
- 🧠 AI / ML Engineering — multi-agent orchestration, RAG pipelines, hybrid retrieval, LLM evaluation
- 🏗️ Full Stack Engineering — Next.js, React, Node.js, FastAPI, Django production systems
- ⚙️ Product Engineering — shipping systems with CI/CD, observability, and measurable performance impact
- ☁️ CRM / SaaS Engineering — Salesforce Apex, SOQL, LWC, workflow automation
Open To:
Languages
Frontend
Backend & Databases
Cloud, DevOps & Tooling
| Domain | Proficiency | Details |
|---|---|---|
| Multi-Agent Orchestration | ●●●●○ | LangGraph state machines, agent routing, deterministic verification gates |
| Retrieval-Augmented Generation | ●●●●○ | Hybrid BM25 + dense vector retrieval, pgvector, section-aware chunking |
| LLM Integration | ●●●●○ | OpenAI GPT-4o / GPT-4 Turbo, prompt engineering, structured JSON outputs |
| ML Frameworks | ●●●○○ | PyTorch, TensorFlow, scikit-learn |
| Evaluation & Observability | ●●●●○ | RAGAS metrics, CI evaluation gates, per-query cost/latency tracing |
| Generative AI Deployment | ●●●○○ | OCI Generative AI workflows, LLM deployment concepts |
🔹 FilingSense — Multi-Agent RAG Platform for Financial Due Diligence
Multi-agent, retrieval-augmented system for financial due diligence over SEC 10-K/10-Q filings. Routes analyst questions to specialist agents, retrieves evidence with hybrid search, and verifies every answer against source text before returning it.
| Stack | Python · FastAPI · LangGraph · PostgreSQL · pgvector · OpenAI GPT-4o · Streamlit · Docker · GitHub Actions · RAGAS |
| Scale | 1000+ filing sections ingested via automated, section-aware chunking pipeline |
| Performance | +35% query routing accuracy via LangGraph multi-agent orchestration |
| Security | Deterministic verification engine — 100% citation-backed responses |
| Impact | −90% AI hallucination rate, +40% evidence retrieval precision via hybrid BM25 + pgvector search |
| Repository | View Source |
Built a LangGraph-powered orchestrator coordinating three specialized agents (table QA, risk-diff, general grounding) behind a single API. A deterministic verifier gates every response against retrieved chunks before it reaches the user, eliminating ungrounded answers rather than merely reducing them.
🔹 ConvoGenius — AI-Powered Video Conferencing PWA
Mobile-first, AI-powered video conferencing progressive web app built to boost meeting productivity through live transcription and real-time collaboration.
| Stack | Next.js · TypeScript · tRPC · Drizzle ORM · PostgreSQL · Stream Video & Chat SDKs · OpenAI GPT-4 Turbo · Inngest · Vercel CI/CD |
| Scale | 5 Drizzle ORM schemas across meeting and user records, fully ACID-compliant |
| Performance | <200ms Time-to-Interactive, 99.8% uptime, 30% reduction in global call latency |
| Security | ACID-compliant schema migrations — 40% reduction in data errors |
| Impact | +25% user engagement, 95% transcription accuracy with speaker diarization, summaries generated in <60s |
| Repository | hacksummit-22wy.vercel.app |
Delivered HD video calling via Stream Video & Chat SDKs, layered with live GPT-4 Turbo transcription and speaker diarization, and shipped through a full Vercel CI/CD pipeline.
Salesforce Developer Intern · TechForce Academy (SRMIST)
Aug 2025 — Nov 2025
Contributed to CRM platform engineering on the Salesforce ecosystem, automating business logic and streamlining data operations across custom and standard objects.
- Developed Apex triggers and classes to automate business logic, reducing manual processing effort
- Wrote SOQL queries to efficiently retrieve and manipulate CRM data across Salesforce objects
- Configured CRM workflows, validation rules, and process automation to meet business requirements
- Gained hands-on experience with the Salesforce Lightning platform and SaaS-based cloud CRM development
| Recognition | Details |
|---|---|
| 🏆 1st Place — CODECRUX: JavaSprint, Innovatex'25 | Organized by SRMIST ACM SIGAI Student Chapter (Apr 21, 2025) — built "Study Habit Tracker," a Java application increasing user productivity by 40%, using Java, Swing, Google Gemini API, MySQL, and JFreeChart |
Oracle
Generative AI workflows, prompt engineering, OCI AI services, LLM deployment concepts
Cloud fundamentals, OCI core services, networking, storage, security, IAM, compute, and cloud architecture concepts
CI/CD pipelines, DevOps automation, Infrastructure as Code, Kubernetes, monitoring, logging, and OCI DevOps services
Salesforce
Agentforce AI, Prompt Builder, Copilot, Data Cloud, AI-powered CRM workflows, and Salesforce automation
TechForce Academy
Apex triggers, SOQL, Salesforce CRM development, Lightning Web Components, SaaS architecture
current_focus:
learning:
- Advanced multi-agent LLM orchestration (LangGraph, MCP)
- CI/CD pipelines and container orchestration
- Production-grade RAG evaluation methodologies
building:
- FilingSense — multi-agent RAG for financial due diligence
- Full-stack AI-integrated platforms
exploring:
- Firecrawl-based agentic web retrieval
- Cloud-native deployment patterns
open_to:
- Software Development Engineer internships
- AI / ML Engineering roles
- Open source collaboration
"Code is the bridge between ambition and impact — I build systems that hold up under scrutiny."