Forward Deployed Engineer · Melbourne · Open to work

Hi, I'm Manav.

I build AI and autonomous systems and stay with them until they work in production. Voice agents that called real customers, clinical tooling doctors used, and peer reviewed research along the way. Currently in Melbourne, reading for a Master of Data Science.

See the work →works.imanav@gmail.com

GitHub LinkedIn Medium Scholar Resume

01 / Experience

Where I've worked

From data science internships to owning AI products in production. Drag the bar or scroll sideways to move through it.

2026 – Present

The Solutionist Agenda ↗

Consultant

Melbourne · Remote

Designing and building thesolutionistagenda.com end to end, from the public site through to the authenticated product.
The platform turns an unresolved business decision into a documented plan with owners and dates attached.

Next.jsLLM productStripe

May 2026 – Present

Atlas Counsel

AI & Web Consultant

Melbourne · Remote

Advising a legal consultancy on where AI is worth deploying, then building and maintaining the tooling and web systems that follow.

LLMsWeb

Jun 2025 – Dec 2025

Square Yards

AI/ML Engineer

Mumbai, India

Built a Voice Agent Studio that takes a written script to a deployable outbound AI caller in under ten minutes, with automated testing and prompt refinement built in.
Delivered a multilingual outbound caller on Twilio, Gemini and ElevenLabs serving property leads across several Indian languages.
Introduced structured post call reporting that the operations team used to tune campaigns and prioritise follow ups.

TwilioGeminiElevenLabsWebSockets

Dec 2024 – May 2025

Nihin Media K.K.

AI Engineer, Intern

Tokyo, Japan · Remote

Developed a radiology assistant that reviews imaging, symptoms and patient history together and returns likely diagnoses with the clinical reasoning stated.
Reduced hallucination by grounding every response in traceable sources across Japanese medical literature, PubMed and PMC.
Benchmarked embedding and retrieval configurations to improve accuracy on medical queries.

Multi-agentRetrievalHealthcare AI

Dec 2024 – Jan 2025

DisruptiveNext

Generative AI Engineer, Intern

Pune, India · Remote

Contributed to early stage LLM and agent based product work for client engagements.

LLMsAI Agents

Sep 2023 – Jan 2024

C1X Inc.

Software Engineer, Intern

Chennai, India

Built content based and BERT based recommendation systems covering 1,100 products and 25,000 users.
Improved recommendation precision with BERT embeddings on AWS, lifting user engagement by 8 percent.

BERTAWS

Jun 2023 – Jul 2023

Capabl

Data Science Intern

Bangalore, India · Remote

Developed and evaluated machine learning models across datasets of 50,000 rows, improving accuracy by up to 15 percent through preprocessing and feature engineering.

Pythonscikit-learn

02 / Projects

Things I built on my own time

LLM · FastAPI

LeaseFlow AI

Extracts structured data from rental agreements and generates tenant welcome packs automatically.

Diffusion

ObjectEraserSD

Removes a chosen object by steering Stable Diffusion's attention maps mid generation, leaving the rest of the scene intact.

Agents

Proposal Generator

A multi agent pipeline that drafts complete business proposals without human intervention.

Vision · Audio

LipSync

Syncs any audio track to video footage with Wav2Lip, including partial or intermittent facial footage.

Vector Engine

Semantic search combining transformer embeddings with Faiss for fast similarity lookup.

Forecasting

Demand Forecasting

Demand and inventory models for e-commerce stocking decisions.

Recently pushed

All repos →

h-mdatasetfiltered dataset of h&m recommendations kaggle challenge.5★Graph4RecommendationsA graph based movie recommender appJupyter Notebook · 1★ConvoContextA question answering model made using transformers.Python · 1★portfoliomy portfolioTypeScript

03 / Research

Published research

2025 · Springer LNNS · ICIVC

FasterRes-FPN: A Deep Learning Approach for Microplastic Pollution Surveillance in Water Bodies

Faster R-CNN with a ResNet-50-FPN backbone, reaching an F1 of 0.9376 on real aquatic data and outperforming the YOLOv8 and YOLOv9 baselines.

Israni, Biswas, Renuse, Premanand 2025 · IEEE · ICDSAAI

AutoMed: Multi-Agent AI System for Personalized Medical Knowledge Retrieval and Summarization

Agents that refine the query, retrieve the literature and curate the answer for biomedical questions, at roughly four seconds to first token.

Israni, Renuse 2025 · Springer LNNS · ITAI

RCA-MViT: Residual Convolutional Attention Multi-scale Vision Transformer for Dermatological Disease Classification

ResNet18 combined with CrossViT for multi-class skin disease classification. 90.5 percent accuracy, ROC-AUC 0.9879.

Israni, Premanand, Renuse

04 / Education

Where I studied

Studying now2026 – 2028

The University of Melbourne

Master of Data Science

Melbourne, Australia

Graduated2021 – 2025

VIT Chennai

B.Tech in Computer Science

CGPA 8.92 out of 10

Chennai, India

Also attendedAmazon ML Summer School, 2024ACM India Summer School, IIT Gandhinagar, 2024

05 / Off the clock

A little more of me

I write, I get outside, and I spend a lot of time meeting people who build things. If you want a second pair of eyes on something, I am easy to reach.

Medium↗

Essays on LLM tooling and voice agents

Anthropic Priced the Danger: What Claude Fable 5 Really Tells Us About AI Safety
How 250 Bad Files Can Hack a Billion-Parameter AI
The Secret to Reliable Voice Agents: Treat Prompts Like State Machines

Read on Medium

Substack↗