~/now-loading bhumika_marmat

BHUMIKA MARMAT tashasho

I build the tools VCs wish they had. Across funds in Palo Alto, Bengaluru, and the Japan–India corridor — production code that runs in live IC environments, not slide decks.

role:
Pipeline +42%
Automation 83%
Startups 3,500+
Alpha $67M
▼ Download CV @ Email ▣ Book a Chat in/ LinkedIn gh/ GitHub ✎ Substack 
def alpha(fund):
  return bayes(fund)
  // +$67M
while reading:
  patterns++
commit -m "VC ≠ vibes"
Python Monte Carlo Bayesian Inference FastAPI React TypeScript Multi-Agent AI Affinity CRM NLP Transformers XGBoost Game Theory Stochastic Processes Next.js Supabase Zapier n8n Gemini AI Python Monte Carlo Bayesian Inference FastAPI React TypeScript Multi-Agent AI Affinity CRM NLP Transformers XGBoost Game Theory Stochastic Processes Next.js Supabase Zapier n8n Gemini AI
// 01 — work_history

QUEST
LOG

Three funds, three stacks, one obsession: why investment decisions so rarely match the mathematics of uncertainty.

QUEST COMPLETE ✓

Venture Capital Intern

@ Z5 Capital · Palo Alto, CA

Aug 2025 – Apr 2026 Remote / Palo Alto

Built Nexus-9 AI underwriting engine and automated deal-flow for a $50M Palo Alto fund.

  • Architected Nexus-9, a 4-agent AI underwriting engine deployed in live IC deal review — Monte Carlo stress-testing (n=10,000), Black Swan scenarios, Bayesian founder-bias correction, and Runway AI psychometric games (r = 0.38–0.61 with deal outcomes).
  • Engineered end-to-end deal-flow automation: IMAP stealth-sourcing parser, Affinity CRM relational mapping, IC memo auto-generation, and SEC Filings Analyzer — drove +42% sourced-to-pipeline conversion for a $50M fund.
  • Led 150+ founder calls across 368 sourced startups; pushed 17 to dealflow pipeline and supported 2 SPV launches.
PythonMulti-agent AIMonte CarloAffinity CRMFastAPI
QUEST COMPLETE ✓

Venture Capital · Beyond Next Ventures

@ beyondnextventures.com · Bengaluru

May – Jul 2025 Bengaluru

Cut analyst review time 83% for a ¥50B Japan–India deep-tech fund.

  • Automated inbound sourcing across Affinity CRM, Python & Zapier — cut analyst review time 83% daily, unblocking evaluation of 3,500+ startups for a ¥50B Japan–India fund.
  • Led market sizing for 6 deep-tech targets; built co-investor pipeline of 118 micro-VCs for Japan→India deal syndication.
ZapierAffinityn8nJapan–India corridor
QUEST COMPLETE ✓

Associate

@ Grad Capital · Bengaluru

Oct – Nov 2024 Bengaluru / Remote

Sourced 200+ startups, pushed 8 to pipeline, 2 funded.

  • Sourced 200+ startups across IIT KGP & Ashoka; led 35+ founder calls — 8 pushed to pipeline, 2 funded. Built Virtual Data Room improving efficiency by 30%.
Deal SourcingPitch ReportsData Room
// 02 — built_things/

~/CRAFTED/tools

Production code, not portfolio code. Each one runs in real environments — IC decks, fund pipelines, founder workflows.

T-001

Nexus-9

4-agent AI underwriting engine for live IC deal review.

Monte Carlo stress-testing, Black Swan scenario modelling, Bayesian founder-bias correction, and psychometric game-based founder assessment.

TVPI: 1.24× → 1.91× · 54% reduction in Type II errors
TypeScriptMulti-agentMonte CarloReact
↗ GitHub
T-002

Deal-Flow Engine

Enterprise AI deal-flow pipeline processing 10-20 deals/day.

Multi-channel sourcing from LinkedIn, GitHub Trending, Twitter/X, Product Hunt, Hacker News, Reddit. Gemini 2.0 scoring, Slack-native triage, and automated cold outreach.

+42% sourced-to-pipeline conversion · 10–20 deals/day
PythonFastAPIGemini AISlackAirtable
↗ GitHub
T-003

SEC Filings Analyzer

Automated SEC Form D parsing for investment insights.

Parses SEC Form D filings to surface investment landscape insights. Extracts NDR trends, revenue multiples, and M&A triggers from S-1s and 10-Ks at a fraction of analyst time.

Automated extraction across S-1 / 10-K / Form D corpus
PythonNLPTransformersPlotly
↗ GitHub
T-004

Inbound Sourcing Workflow

Automated inbound evaluation for a ¥50B fund.

Automated inbound deal evaluation integrating Affinity CRM, Python, and Zapier. Cut analyst review time 83% daily at a ¥50B Japan–India fund, enabling coverage of 3,500+ startups.

83% analyst time reduction · 3,500+ startups evaluated
ZapierAffinityPythonn8n
↗ GitHub
T-005

Startup Investments Patterns

Crunchbase data analysis for investment patterns.

Data analysis of Crunchbase investment data (2013) to understand why certain investments worked or didn't. Pattern recognition across funding rounds, sectors, and outcomes.

Crunchbase pattern analysis
PythonData AnalysisCrunchbase
↗ GitHub
S-001

Forge

AI resource orchestration for early-stage founders.

AI-powered resource orchestration system for founders. Helps early-stage teams find the right resources, connections, and operational support. Built on Next.js with Supabase backend.

Full-stack founder tooling · Next.js + Supabase
TypeScriptNext.jsSupabaseReact
↗ GitHub
S-002

thinkor

Thinking-out-loud companion for Claude prompting.

A thinking-out-loud companion for Claude prompting. Helps iterate on prompts through structured conversation and reflection — because the gap between what you want and what you ask for is where most AI value leaks.

Claude prompting assistant
PythonLLMPrompting
↗ GitHub
S-003

Social Media Automator

End-to-end social posting automation for X + LinkedIn.

End-to-end social media posting automation for X and LinkedIn. Handles the full content pipeline from creation to posting so you stop wasting time on 1mm alignments in Canva.

Automated X + LinkedIn posting pipeline
PythonAPIAutomation
↗ GitHub
X-001

CityScope

Stardew Valley-style pixel art event maps for Indian cities.

Zero-dependency, pure HTML5 Canvas at 60fps. Renders live events across Bengaluru and Mumbai as a fully interactive retro game world with auto-rickshaws, monsoon rain, and metro trains.

Zero dependencies · Pure Canvas · 75+ live events
HTML5 CanvasJavaScriptPixel Art
↗ GitHub ▶ Live Demo
// 03 — papers/

LIBRARY
{ research }

Where the math actually lives — Bayesian frameworks, stochastic processes, signalling games.

PAPER 01

A Probabilistic & AI-Augmented Framework for Early-Stage VC

Prof. T. Bhattacharjee · IIT Kharagpur · BTP · 2025–2026

Developed the Bias-Adjusted Valuation (BAV) model — a Bayesian odds-ratio framework formalising cognitive biases in VC underwriting, calibrated on 500+ Indian deep-tech deals. Extended to a dynamic Ornstein–Uhlenbeck process with signalling-game equilibrium, proving a chilling effect on patent disclosure under information asymmetry. Copula-based portfolio extension revealed severe tail-risk underestimation under independence assumptions.

β_geo ≈ 0.22 $67M alpha / $100M fund 2.4× tail-risk ↑ 54% ↓ Type II errors TVPI 1.24× → 1.91×
PAPER 02

Overcoming Subjectivity Bias by Feature Modelling

Prof. T. Bhattacharjee · IIT Kharagpur · 2025

SVM RBF regression predicting startup valuation across 15+ features, optimised with sigmoid threshold tuning for VC risk-aligned readiness scoring.

>80% valuation accuracy F1 = 0.77 synthetic
PAPER 03

Game-Theoretic Optimisation on Corruption Networks

Prof. Dripto Bakshi · IIT Kharagpur · 2024

2-way mathematical optimisation (path & cost) on political social networks modelled as directed graphs. 48% success rate via reward-based routing on large organisational DAGs.

48% routing success Directed graph DAGs
// 04 — substack/

notes.md

Six pieces about why investors get it wrong — and what founders are actually thinking.

NOTE 01

Your VC Portfolio Model Is Lying to You About Risk

How geographic bias distorts deal decisions and changes founder behaviour through the chilling effect.

 NOTE 02

Why Biased Investors Make Founders Worse

On the Bias-Adjusted Valuation model and geographic pessimism in Indian deep-tech investing.

 NOTE 03

The Companies That Should Be Funded But Aren't

A pattern I kept running into while evaluating early-stage deep-tech companies.

 NOTE 04

What the Room Is Actually Thinking When You Walk In

Nobody tells you this part.

 NOTE 05

The Traits That Actually Predict Whether You'll Still Be Building in Two Years

After sitting in on 150+ founder calls across early-stage deals in India and the US.

 NOTE 06

The Anxious Founder Who Was Right on Time

Something the startup world doesn't say out loud very often.
↗ Read all on Substack
// 05 — side/

SIDE
QUESTS

▣ cityscope

The whole city as a game.

Pixel-art live event maps for Indian cities — zero dependencies, pure HTML5 Canvas at 60fps. Auto-rickshaws, monsoon rain, metro trains. 75+ real events rendered as a retro game world.

↗ github ▶ live demo
▣ thinkor

The gap between what you mean and what you ask is where the value leaks.

Thinking-out-loud companion for Claude prompting. Iterate on prompts through structured conversation and reflection — close the gap before you ship the request.

↗ github
▣ lvl-up

Build yourself like a JRPG character.

Personal-growth tracker — habits, skills, XP, milestones. Discipline as gameplay. Daily quests, weekly bosses, gear that actually changes your stats.

↗ github
▣ forge

What founders actually need, before they know they're looking.

AI resource orchestration for early-stage teams. The right intros, the right ops support, the right templates — surfaced as the company moves. Next.js + Supabase.

↗ github
▣ social-media-automator

Stop wasting time on 1mm alignments in Canva.

End-to-end posting pipeline for X and LinkedIn — draft, schedule, post, archive. So the writing happens once and shipping happens automatically.

↗ github

~/brewing

Forks, side soups, and half-cooked experiments. Things I'm pulling apart, contributing to, or stewing on.

⑂ all forks → ★ starred → everything

scholar → NTSE National Scholar 2020 — top 0.1% of 1M+ candidates · NCERT Government Scholarship · INSPIRE Fellowship candidate (Govt. of India, top 1% of national exams).

case-comps → National Runner-Up, MnAnalyse M&A Case Competition (IIT Guwahati, LBO — Zomato × Blinkit) · National Finalist, Insight Masters Business Analytics (ISB Hyderabad) · National Finalist, AWL Resolve (Adani Wilmar × IIM Ahmedabad).

discourse → Silver Contingent, ACWG Debate (G20). Winner IIT KGP Intras. Broke at LSRPD & DTU PD. Best Debater, Sahodaya Regional · Public Speaking Recognition.

founding → Founded the Ujjain chapter of Period, a global menstrual-health org. Volunteer, Robinhood Army & Robinhood Academy.

writing → Youngest journalist at the Free Press Journal.

hardware → Silver, Aavishkar International Robotics Championship · UCMAS National Merit.

music → 2× Gold Anuvrat National Singing · Gold Anuvrat State Art · Merit Gandharva Mahavidyalaya · currently strumming ukulele 🎵

// 06 — about_me.txt

the player behind
the profile.

I'm a final-year Mathematics & Computing student at IIT Kharagpur, graduating in 2026. I ended up in VC not by accident but by chasing a specific question: why does the way investment decisions get made so rarely reflect the actual mathematics of uncertainty?

That question has taken me across three funds — Z5 Capital in Palo Alto, Beyond Next Ventures in Bengaluru, and Grad Capital — and through a year of supervised research with Prof. Bhattacharjee formalising those intuitions into working models. The tools I've built aren't portfolio projects; they run in live IC environments.

I'm looking for a full-time analyst seat at an early-stage fund — ideally deep-tech, AI, or India corridor — where the technical depth I bring actually changes how investment decisions get made. Based in Bengaluru, open across India.

Outside investing: I debate (broke at LSRPD, G20 Silver), founded a social chapter in Ujjain, was the youngest journalist at the Free Press Journal, and I write on Substack. I think clearly in public.

// 07 — connect/

$ ./connect
--me

If you're building that team, I'd love to talk.

Currently in Bengaluru. Open to roles across India. Best reached by email or LinkedIn.

tashashoms@gmail.com EMAIL linkedin.com/in/bhumikamarmat LINKEDIN github.com/tashasho GITHUB substack.com/@bhumikamarmat SUBSTACK Book a Chat CAL