VJ
№ 001 — Portfolio / 2026
Vadodara · 22.307° N, 73.181° E
Updated 04 · 2026

Vatsa Joshi.
building agents that ship.

Full-stack engineer working at the seam between large language models, multi-agent systems, and the messy production infrastructure that has to carry them. Currently shipping legal AI at Promact.

RoleSoftware Engineer (AI)
FocusAgentic systems · RAG · LLM fine-tuning
StackPython · TypeScript · Rust
BasedVadodara, Gujarat (IN)
Résumé See the work Say hi
Orchestrator
vatsa.agent
status: idle · 04:12 UTC
Stream
> boot orchestrator.py ...
Context
tools17tokens1.2Mruns4,318uptime99.2%
01 — ABOUT

The human
behind the agents.

Vatsa Joshi
22 · India
EXHIBIT AVatsa Joshi speaking at Navrachana University
↳ speaking @ Navrachana University, 2025

I build end-to-endsoftware — from low-level model implementations to production full-stack systems — with a founder's bias for shipping.

Computer science undergraduate at Navrachana University, graduating September 2026. I've spent the last three years working across the AI/ML and SaaS industries as an intern, contractor, and now a full-time trainee — specializing in agentic systems, retrieval pipelines, and the boring-but-critical infrastructure that makes LLMs behave in production.

The stuff I actually care about: multi-agent architectures with real coordination logic, RAG systems that don't hallucinate, fine-tuning small models for specific domains, and writing transformers from scratch in Rust when nobody asked me to.

DegreeB.Tech Computer Science
Navrachana University · 2022–2026
InterestsAgents, systems, low-level ML, research
ReadingDistributed systems, compiler internals
ShipsGitHub @Vatsa10 · 40+ repos
02 — NOW

What I'm shipping
right now.

Jan 2026 → Present
Promact, Vadodara
Software Engineer
Trainee.
Promact · Vadodara · AI Division

Contributing to BharatLaw.ai — an AI-powered legal technology product serving Indian law firms. Designing the AI systems that read, understand, and draft across millions of legal documents.

  • Built end-to-end AI-driven Document Management Systems — translation, parsing, indexing, retrieval across 22+ languages.
  • Shipped a multi-agent Petition Writer that drafts structured legal petitions from case inputs, with iterative human-in-the-loop clarification before the final draft.
  • Designed a legal intelligence platform with judge profiling, case-brief-driven litigation strategy, and lawyer compatibility analysis — agents coordinating vector similarity, LLM reasoning, and live web search.
Flagship productBharatLaw.ai
CompanyPromact Infotech
RoleSoftware Engineer Trainee (AI)
StartedJanuary 2026
DomainLegal-tech, multi-agent AI
StackPython, FastAPI, LangGraph, Elasticsearch, PostgreSQL
ScopeEnd-to-end AI systems
03 — EXPERIENCE

A short trail
of roles.

2023 → Present
5 positions
2026 — NOW
Software Engineer (AI)Promact · Vadodara
BharatLaw.ai — multi-agent petition drafting, legal intelligence platform, end-to-end document systems.
Full-time
Current
JUN — DEC 2025
Full-Stack DeveloperHiraeth Tech · Vadodara
Built scalable products with NestJS, Next.js, PostgreSQL; integrated LLM-based services and vector search into the web stack.
Internship
6 months
MAR — MAY 2025
Deep Learning EngineerProdigal AI · Remote
CNNs for image classification (92% accuracy), fine-tuned transformer models, real-time object detection with YOLO / SSD.
Internship
3 months
FEB — APR 2025
AI / ML EngineerSparks To Ideas · Remote
XGBoost churn prediction at 95% accuracy; built an AI content intelligence pipeline with FAISS (HNSW) deduplication using multilingual embeddings.
Internship
3 months
MAY — JUL 2024
Summer Intern, BackendPAM Infotech · Vadodara
Java backend work, legacy migration to modern stack, server-side performance optimization.
Internship
3 months
MAY — JUL 2023
Web Application DeveloperJeenweb Technologists
First role — React frontends, backend integration, end-to-end web apps. Where it all started.
Internship
3 months
04 — PROJECTS

Things I've built
on purpose.

Selected work
2023 → 2026
P · 01Agents
Conversational multi-agent appointment booking — real-time availability, automated email confirmations. Orchestrator + DB + Notifier.
Booky
Multi-AgentFlaskSQLite
P · 02Agents
N8N-inspired agentic workflow automation with a visual node builder and AI-powered conditional routing between agents.
FloWork
LangGraphGroqNext.js
P · 03Agents
Agentic recruitment chatbot — candidate screening, dynamic technical assessment generation, sentiment-based evaluation.
Hiring Assistant
CrewAIGemini 2.0Streamlit
P · 04Agents
Personal AI agent automating Gmail, Google Calendar, and Tasks via Workspace APIs with OAuth 2.1 and real-time chat.
PersonAI
GroqLangChainFastMCP
P · 05Agents
Multi-agent WhatsApp shopping assistant for product discovery, price comparison, and informed purchase decisions.
Ecommerce Agent
LangGraphLLMWhatsApp
P · 06RAG & Memory
Enterprise legal document parser — 22+ language OCR, RAG-ready semantic chunking, and Elasticsearch indexing.
LegalOCR
PaddleOCRFastAPIElasticsearch
P · 07RAG & Memory
Multi-channel AI memory unifying conversations across web, WhatsApp, email, and phone with probabilistic identity linking.
Synapse
Next.jsQdrantUpstash
P · 08RAG & Memory
Multi-agent system converting financial documents — 10-K / 10-Q, SEC filings, earnings transcripts — into actionable intelligence.
Finance Advisor
RAGMulti-AgentLLM
P · 09RAG & Memory
Voice-first AI email assistant — read, summarize, star, and compose your inbox via natural conversation. Hands-free.
Email Voice AI
LangGraphDeepgramGmail API
P · 10Fine-tuning
Fine-tuned Qwen3.5-0.8B on 1.5k Hindi → Gujarati pairs using ZeRO-2/3 multi-GPU distributed training.
Hindi2Guj
Qwen3.5DeepSpeedAccelerate
P · 11Fine-tuning
Fine-tuned Gemma 3 (4B) on FineTome-100k for enhanced instruction-following via LoRA / QLoRA on a single Colab T4.
Gemma 3 Tune
UnslothLoRAHuggingFace
P · 12Fine-tuning
Complete transformer architecture implemented from scratch in Rust, for performance-focused inference research.
Rustformer
RustTransformerFrom Scratch
P · 13Vision / Mobile
2nd place at Tinkerthon 3.0 + 1-year SSIP Gujarat research grant. Fine-tuned DL models for deepfake video classification.
Deepfake Detector
EfficientNetB7BiLSTMCV
P · 14Vision / Mobile
Mobile health app that translates lab reports into plain language — with fully on-device medical transformer inference.
AarogyaKosh
React NativeFastAPIOn-device
05 — STACK

Tools of the trade.

Working knowledge
Updated 2026

Languages

  • Python
  • TypeScript
  • Rust
  • Java
  • C

AI / ML

  • PyTorch
  • TensorFlow
  • HuggingFace
  • Unsloth
  • Scikit-learn

LLM & Agents

  • LangChain
  • LangGraph
  • CrewAI
  • RAG / CAG
  • LoRA · QLoRA
  • FAISS · Groq

Backend & Cloud

  • FastAPI
  • NestJS
  • Node.js
  • Docker
  • AWS
  • PostgreSQL · Mongo · ES

Frontend

  • Next.js
  • React
  • React Native
  • Tailwind

Vision

  • OpenCV
  • YOLO / SSD
  • PaddleOCR
  • Diffusers

Data & Infra

  • Elasticsearch
  • Qdrant
  • Upstash Redis
  • Inngest

Practices

  • End-to-end ownership
  • Human-in-the-loop
  • Shipping > perfecting
06 — AWARDS / EDU / PUBS

Honours & paper trail.

2020 → Present

Tinkerthon 3.0

Second place for a Deepfake Video Detection System. Fine-tuned DL pipeline selected among top hackathon submissions.

Feb 2025 · Hackathon
§

SSIP Gujarat Grant

One-year research grant from SSIP Gujarat to further develop the deepfake detection project into a production-ready tool.

2025 · Research funding

Odoo × NMIT Hackathon

Top 5 finish out of 3,300 competing teams — rapid product design and implementation against a shared brief.

2024 · Hackathon

B.Tech · Computer Science

Navrachana University, Vadodara. Specialisation in Software Engineering & AI. Research on quantum-dot memory, EfficientNetB7 + BiLSTM for deepfake detection, and stable-diffusion VAEs.

2022 → Sep 2026

Rabin-Karp Algorithm

Report & analysis of the Rabin-Karp string-matching algorithm. Published on ResearchGate.
DOI: 10.13140/RG.2.2.24592.01283

Publication · ResearchGate

Blockchain Technology

Survey paper exploring blockchain architecture, cryptographic foundations, and cross-industry applications.
DOI: 10.13140/RG.2.2.32465.75365

Publication · ResearchGate

Self-Adaptive Context RAG

Invited talk on a RAG architecture that tunes its retrieval context on the fly — delivered to the Google Developers Group community.

Talk · Google Developers Group

Agentic AI

Gave a workshop on multi-agent systems, tool-use patterns, and shipping agentic AI into production. Navrachana University.

Workshop · Navrachana University
07 — CONTACT

.

Open to: full-time · collab
Reply time ~24h

Let's build something that ships.

E-mailvatsajoshi2@gmail.comPhone+91 97375 57799GitHub@Vatsa10 ↗Webvatsajoshi.in ↗