Tata Institute, Mumbai
MSc Analytics
Atharva R.
Avid hobby picker-upper, based in Mumbai.
Education
Narsee Monjee College, Mumbai
BSc Information Technology
Work
Aritra Partners
Mumbai, India
-
AI/BI Manager Lead, Strategic Initiatives
Built a complete due diligence interface that takes the team from the first data collection to a finished report.
Built Ari, the firm's governed AI portal for 60+ staff, with lower costs, usage auditing, and defined outputs.
Ran AI fluency training across the firm and for VC partners, and built productivity and transcription tools for the team.
-
AI/BI Senior Associate
-
AI/BI Associate
Built SecDesk, the approval platform Blume Ventures runs its portfolio consents on. 8+ VC partners use it, and it saved 85% of two compliance executives' time.
Shipped FMV Valuation and Verity, products for valuations and term sheets, cutting report turnaround from days to hours.
Turned the firm's brand into reusable AI skills: logo system, palette, and typography.
Projects & Research
Shiny Dex (100%) in Pokémon FireRed (Nintendo Switch), using Applied Math
Obtained all 182 Pokémon in FireRed, plus all 28 Unown forms. Each one was caught or hatched shiny on the original Switch release, with no emulators or cheats.
Used RNG manipulation to time button presses to the exact 16 ms frame with EonTimer and planned target frames with community tools (Ten Lines/PokeFinder). Recalibrated after each attempt to find IVs from the stats of test catches.
Pulled off odds of up to 1 in 50 million, including a shiny Dragonair holding a Dragon Fang (never found before) and a shiny Chansey holding a Lucky Egg. Learned the game's random-number mechanics straight from its decompiled source code.
stack EonTimer, Ten Lines, PokeFinder, the pokefirered decompilation
hardest bug TBF
note TBF
Memes’ impact on concept simplification in STEM
MSc dissertation, Tata Institute
Studied AI-transformed video memes: familiar internet personalities, voice-cloned, explaining science, maths, and engineering to people who came for the joke. The familiar face is nostalgia marketing at work, and the concept rides in on it.
Proposed the Parasocial Trojan Horse: the creator is the vehicle, the concept is the payload, and the viewer opens the gate for entertainment, not for a lesson. Read thousands of comments across hundreds of Instagram posts, and found curiosity spilling over far more often than criticism.
stack Python, VADER, regex pattern detection, a hand-built Instagram comment corpus
hardest bug TBF
note TBF
Image Synthesis
Using Stable Diffusion
Latent diffusion starts from pure noise in a compressed latent space and removes it step by step, guided by a text prompt, until a decoder turns what remains into pixels.
Built a working text-to-image pipeline on Stable Diffusion, from prompt to finished image.
stack Stable Diffusion, Python
hardest bug TBF
note TBF